LLM 관련 주요 논문 - 2026-06-29

1. Tandem Reinforcement Learning with Verifiable Rewards

Authors: Difan Jiao , Raghav Singhal , Robert West , Ashton Anderson
URL: https://arxiv.org/abs/2606.28166
Abstract:

Reinforcement learning with verifiable rewards (RLVR) has significantly improved the reasoning capability of large language models, reaching expert or even superhuman performance in domains such as competition math. However, whether weaker agents and humans can actually harness this capability is far less certain, with RLVR documented to drift reasoning toward idiosyncratic patterns such as poor readability and language mixing. Tandem training is a recently introduced paradigm that targets this compatibility problem: a trained, stronger senior co-generates each rollout with a frozen, weaker junior, and the two are rewarded as a team, so the senior is pushed to reason in ways the junior can follow. Yet this paradigm has so far been demonstrated only in proof-of-concept settings, leaving open whether it scales to the long chains of thought of the modern RLVR pipeline. In this work, we propose Tandem Reinforcement Learning (TRL), which carries the tandem training paradigm into RLVR. In TRL, the senior and a frozen junior alternate stochastically to co-generate the reasoning, the resulting generation is rewarded, and the standard GRPO loss is applied to the senior. Training Qwen3-4B-Instruct on competition math, we find that TRL matches vanilla GRPO on solo reasoning capability while three properties emerge together from the same rollout structure: stronger handoff robustness with the junior, reduced distributional drift from the junior, and a chain-of-thought more legible to the junior. Our results demonstrate a promising route for RLVR with practical payoffs in multi-model communication and human compatibility.

2. JD Oxygen AI Item Center (Oxygen AIIC) V1: An Industrial-Scale LLM/VLM-Centric Solution for Item Understanding, Management, and Applications

Authors: Oxygen AIIC , Chan Long , Chao Liu , Chaofan Chen , Chaohui Dong , Chunyuan Guo , Danping Liu , Debin Liu , Deping Xiang , Fulai Xu , Guangyue Liu , Hao Li , Huichun Hu , Jian Yang , Jianan Wang , Jianbo Zhao , Jiaoyang Li , Jiaxing Wang , Jinglong Li , Jinjin Guo , Jun Fang , Jun Liu , Kai Zhou , Li Wang , Lili Gao , Liying Chen , Luning Yang , Mengdi Zhou , Pengzhang Liu , Qi Lv , Qianyun Wang , Qixia Jiang , Ruyue Li , Shimu Liang , Shuxing Wang , Sijie Zhang , Siqi Li , Tianhao Gao , Wang Ke , Weihu Huang , Wencan Lai , Wenjie Zhang , Xiaohui Zhang , Xiaojing Dong , Ya Liu , Yifeng Zhang , Yixiang Wang , Yongtai Zhang , Yongyi Liao , Zhaoru Chen , Zhen Chen , Zhiyong Ma , Zhiyuan Liu , Zhongwei Liu , Ziyan Xing
URL: https://arxiv.org/abs/2606.28070
Abstract:

this http URL , one of the world’s largest e-commerce platforms, serves over 700 million active users and millions of merchants, with a catalog of tens of billions of SKUs. At this scale, high-quality, structured item knowledge underpins a better consumer experience, lower management costs, and higher operational efficiency-yet producing and serving it poses three industrial-scale challenges: fast-emerging concepts, high-quality knowledge production for massive SKUs, and diverse downstream requirements. To address these challenges, we present the JD Oxygen AI Item Center (Oxygen AIIC), an industrial-scale platform built on LLMs/VLMs for item-knowledge production and service. Oxygen AIIC is built around four core pillars: (i) ontology engineering driven by efficient human-AI collaboration, which supports the dynamic evolution and agile expansion of an ontology with millions of entries; (ii) a “Semantic Search then Discrimination”(S2D) knowledge identification architecture that, combined with throughput improvement strategies, enables scalable, extensible, and high-throughput AI Item Library production for tens of billions of SKUs; (iii) self-evolving item-understanding LLMs/VLMs that improve in a stable and controllable manner, enabling knowledge production with 94.2% precision and 82.8% recall; and (iv) a unified item tunnel that serves as the data and service hub. Oxygen AIIC now covers tens of thousands of JD categories and processes hundreds of millions of item updates per day on Huawei Ascend NPUs. It has accumulated hundreds of billions of item-knowledge assets. Deployed across core business scenarios-including search, recommendation, operations, category planning-Oxygen AIIC has delivered measurable gains at scale. Search-traffic coverage reaches 80.4%, item-information quality issues drop by 37%, the automated fill rate of core attributes during item listing exceeds 80%.

Authors: Shiyun Zhao , Xinwei Song , Tianyu Guo , Xiaomeng Gao , Mingyuan Liu , Xu Han , Yuanyuan Zhang , Zhenliang Zhang , Xue Feng , Bo Dai
URL: https://arxiv.org/abs/2606.27826
Abstract:

Multimodal large language models (MLLMs) are increasingly deployed as embodied planners in egocentric environments, where task success requires not only achieving instructed goals but also acting in socially appropriate ways. While explicit goals may render certain actions optimal, implicit social norms often impose hidden constraints. Existing evaluations typically focus on explicit goal achievement or direct norm knowledge, seldom assessing whether planners can infer and apply these hidden constraints within action sequences. We introduce NormAct, a benchmark for embodied social-norm interactions that evaluates plans on Goal Achievement, Norm Compliance, and overall Task Success. NormAct uniquely embeds hidden norms within ordinary tasks, testing whether models can realize them without explicit instruction. Experiments with state-of-the-art MLLMs (GPT-5.4, Claude Opus 4.7, Gemini 3 Pro) reveal a significant gap: models achieve explicit goals in 67.3\% of cases, but comply with hidden norms in only 26.4\%. Cue-condition experiments indicate that this gap stems not from a lack of general social knowledge, but from challenges in activating and grounding relevant norms in context. To address this, we propose NormPerceptor, a context-conditioned cue generator that infers scene-relevant norms prior to planning, increasing Task Success from 24.2\% to 46.7\%. Our results underscore the importance of enabling embodied agents to proactively detect hidden norms, ground them in visual evidence, and integrate them as action-planning constraints. Our benchmark is publicly available at this https URL .

4. Grounded Iterative Language Planning: How Parameterized World Models Reduce Hallucination Propagation in LLM Agents

Authors: Xinyuan Song , Zekun Cai
URL: https://arxiv.org/abs/2606.27806
Abstract:

World models for language agents come in two useful forms. An agent-based world model calls an LLM API and reasons flexibly in language, but its errors appear as hallucinated state changes that are hard to score with ordinary regression losses. A parameterized world model is a trained transition predictor; its errors are easier to measure with quantities such as NodeMSE, delta accuracy, and validity accuracy, but it is usually weaker as a standalone planner. We compare these two families on four graph-structured planning benchmarks and introduce operational hallucination metrics for the agent-based case. The comparison motivates \textbf{Grounded Iterative Language Planning} (GILP), which trains only a small parameterized backbone and combines it with API-based agent reasoning. The backbone supplies valid actions, predicted state deltas, risk, and value; the LLM drafts an action and imagined delta; and a consistency gate asks for revision when the two disagree. On real GPT-4o-mini calls, GILP reduces hallucinated-state rate from 0.176 to 0.035. In calibrated simulator ablations, it raises success from 0.668 to 0.838 while adding only ~22% extra LLM calls.

Authors: Jiajing Zhang , Jiamei Jiang , Chenyang Zhang , Feifei Mo , Linjing Li , Daniel Zeng
URL: https://arxiv.org/abs/2606.27757
Abstract:

Large language models (LLMs) have attracted widespread attention from academia and industry, yet their deployment raises critical security concerns regarding robustness and reliability. Planning, a core component of intelligent behavior, remains challenging for LLMs, which often produce infeasible or incorrect solutions in long-horizon decision-making tasks due to inherent complexity. In this paper, we propose a symbolic feedback-driven iterative self-refinement framework to enhance the robustness and reliability of LLMs in long-horizon planning. Specifically, a natural language prompting mechanism is introduced to map logical symbols into natural language descriptions, enabling LLMs to better capture task constraints and semantics. We further design a symbolic verifier that identifies errors and converts them into corrective instructions interpretable by the LLM, thereby guiding self-refinement. In addition, we leverage a plan recognizer to infer goal reachability, facilitating more effective guidance toward desired goals. Empirical results demonstrate that the proposed framework consistently improves both feasibility and correctness in long-horizon planning tasks. This highlights its effectiveness in enhancing the reliability of LLM-based planning and potential to enable more trustworthy AI systems.

6. ToE: A Hierarchical and Explainable Claim Verification Framework with Dynamic Multi-source Evidence Retrieval and Aggregation

Authors: Zhaoqi Wang , Zijian Zhang , Kun Zheng , Zhen Li , Xin Li , Chunlei Li , Jiamou Liu
URL: https://arxiv.org/abs/2606.27736
Abstract:

The rapid spread of fake news poses increasing threats to information ecosystems, especially as AI-generated misinformation under Generative Engine Optimization (GEO) poisoning allows adversarially crafted content to be systematically surfaced by retrieval systems, contaminating LLM reasoning. In this paper, we propose Tree of Evidence (ToE), a hierarchical evidence reasoning framework for automated fact-checking that models each claim as a dynamically expanding argument tree. ToE integrates a reinforcement learning-driven multi-source retrieval agent, an evidence evaluation agent, and an argument tree aggregation algorithm to iteratively decompose, retrieve, and verify claims through an explainable evidence chain. We further provide a theoretical analysis of the retrieval process, deriving a formal error bound that guarantees the learned policy converges to a neighborhood of the information-theoretically optimal policy. Experiments across multiple datasets and backbone LLMs demonstrate that ToE achieves improvements ranging from 4 to 24 percentage points over competitive baselines, with particularly pronounced gains on adversarially poisoned inputs.

7. DysLexLens: A Low-Resource LLM Framework for Analysing Dyslexic Learners Insights from Online Forums

Authors: Dana Rezazadegan , Atie Kia , Phongpadid Nandavong , Dominique Carlon , Jeremy Nguyen , Abhik Banerjee , James Marshall , Anthony McCosker , Yong-Bin Kang
URL: https://arxiv.org/abs/2606.27619
Abstract:

Dyslexic learners increasingly use artificial intelligence (AI) tools to support reading, writing, organisation, and study-related tasks. However, their lived experiences with these tools remain largely underexamined. This paper proposes DysLexLens, a low-resource LLM framework, designed to analyse dyslexic learners experience with AI through online forum discussions. DysLexLens is designed as an end-to-end, evidence-traceable architecture which transforms noisy social media posts into a dictionary-driven corpora, provides knowledge-graph (KG)-based question reasoning, generates verifiable query responses, and enables response evaluation through quantitative and human-grounded assessment. DysLexLens has four key features. First, it employs a dictionary-driven filtering method to construct a more focused Reddit corpus on dyslexia and AI, filtering out noisy and weakly related posts to improve the relevance of data collected from low-resource forum contexts. Second, it integrates LLM-assisted semantic analysis with KG-based query reasoning to uncover meaningful patterns. Third, it has quantitative evaluation metrics (RAGAS and Query Robustness) to measure LLM-generated response performance. Fourth, it provides structured qualitative validation guidelines for assessing response quality, with a specific focus on hallucination and evidence alignment. We demonstrate the effectiveness of DysLexLens using dyslexia-related Reddit forum data and 30 questions. The results show its potential generalisability to other low-resource forum data contexts. DysLexLens, sample data, questions and evaluation results are available at Github to support reproducibility.

8. Odyssey: Constructing Verifiable Local Truth-Preserving Foundation Models

Authors: Sridhar Mahadevan
URL: https://arxiv.org/abs/2606.27593
Abstract:

We introduce a categorical framework called ODYSSEY for constructing verifiable, local truth-preserving foundation models as compositions of foundries: building-block architectural components that specify a cover of local contexts, local representation families, restriction maps, gluing rules, obstruction policies, update obligations, and human-facing views. A foundry is an organized sheaf of knowledge that carries within it an argumentation component. Concrete foundries are built from generic foundries such as evidence/argument, operational decision, institutional/financial, market meaning, scientific challenge, research-program, assistant-build, and evaluation-harness foundries. Universal Foundry Learning (UFL) formalizes foundry construction as a composition of left and right Kan extensions, with left Kan extension rolling local artifacts into candidate foundries and right Kan extension enforcing the restriction, gluing, obstruction, and argumentation conditions required for promotion. Foundry SQL (FSQL) is a small typed query surface for slicing maintained foundry artifacts that uses TICKET (Topos Integration using Causal Kan Extension Transformers) certification for admitting external or pre-built models into durable ODYSSEY state. ODYSSEY is fully implemented and tested across a wide spectrum of concrete foundries, showing that the same categorical machinery supports domain construction, artifact replay, sheaf diagnostics, grounded Toulmin/local-LLM scrutiny, residual-obstruction ledgers, and optimized TICKET-compatible causal-claim extraction across heterogeneous sources. This paper is to be presented as a 2.5 hour tutorial at ICML 2026. The tutorial home page is at this https URL .

9. Internalizing the Future: A Unified Agentic Training Paradigm for World Model Planning

Authors: Xuan Zhang , Zhijian Zhou , Lingfeng Qiao , Yulei Qin , Ke Li , Xing Sun , Xiaoyu Tan , Chao Qu , Yuan Qi
URL: https://arxiv.org/abs/2606.27483
Abstract:

Large language model (LLM) agents have demonstrated strong capability in sequential decision-making, yet they remains fundamentally reactive in long-horizon tasks. Unlike humans who employ “what-if” reasoning to evaluate potential plans before commitment, standard agents lack an internal world model to simulate future outcomes. Therefore, we propose to internalize future-aware planning by training a single autoregressive model to verbalize both a prospective state rollout and a plan-conditioned success estimate-a textual analogue of the Q-value. Crucially, we identify a format-capability gap: simply fine-tuning agents on look-ahead traces during post-training leads to superficial mimicry of foresight without genuine predictive grounding. To bridge this gap, we introduce a three-stage training paradigm: (i) World Model Agentic Mid-Training (WM-AMT) to inject latent predictive capabilities into the policy; (ii) Format-Eliciting SFT (FE-SFT) to structure this injected capability; and (iii) Foresight-Conditioned Reinforcement Learning (FC-RL) to refine the calibration and utility of the generated simulations. Evaluated on search and mathematical reasoning tasks, our approach consistently outperforms other training baselines. Our results demonstrate that effective internal world modeling in LLM agents requires a capability-first training pipeline to achieve grounded and calibrated foresight.

10. When Does Personality Composition Matter for Multi-Agent LLM Teams?

Authors: Aryan Keluskar , Amrita Bhattacharjee , Huan Liu
URL: https://arxiv.org/abs/2606.27443
Abstract:

Personality prompting shapes how large language models communicate, yet whether these behavioral shifts affect objective task outcomes remains under-explored. Prior work shows that agents prompted with low agreeableness produce adversarial language, while those prompted with high agreeableness become cooperative, but the relationship between communication style and task performance has not been systematically examined across multiple domains. In this work, we investigate whether personality composition matters for multi-agent team performance by manipulating personality traits across frontier LLMs on three task domains: structured coding, open-ended research collaboration, and competitive bargaining. We find that personality effects depend critically on task structure. In coding tasks, low agreeableness leads to large communication shifts that have little effect on milestone completion. In open-ended collaboration and bargaining, the same manipulation substantially degrades performance. We discuss implications for multi-agent system design and the limits of personality manipulation.

11. AI-Model Network: Concept, Current State and Future

Authors: Li Zhetao , Zeng Xiyu , Wang Jianhui , Xiao Yong , Liu Zhongren , Wu Junru , Lai Junjie , Huang Jijun , Long Saiqin
URL: https://arxiv.org/abs/2606.27382
Abstract:

While the primary function of computers lies in computation and processing, the core value of the Internet is rooted in sharing and collaboration. Computers create the Internet, and the Internet empowers the value of computers. The rapid development of the Internet, cloud computing, and big data is pushing artificial intelligence into the era of large models (LMs). However, the practical application of LMs is currently hindered by high training costs and deployment complexities, driving a shift toward lightweight, private, and domain-specific models. With the rapid proliferation and wide distribution of heterogeneous models, enabling effective interaction and collaboration among them has emerged as a critical bottleneck that urgently needs to be addressed in LM development. Drawing inspiration from the development of the Internet, this paper proposes the concept, vision, and system architecture of world wide AI-model network (AI-ModelNet). It is a novel paradigm that achieves interconnection, capability sharing, and collaborative reasoning by establishing pathways between models. We first briefly review the current state of single-model and multi-model research. Subsequently, the systemic vision and hierarchical architecture of AI-ModelNet are articulated, followed by validation of the framework’s feasibility through a prototype system and diverse application cases. Finally, key directions for future research are discussed preliminarily.

12. Cognitive Episodes in LLM Reasoning Traces Enable Interpretable Human Item Difficulty Prediction

Authors: Chenguang Wang , Ming Li , Xinyue Zeng , Zhuochun Li , Hong Jiao , Tianyi Zhou , Dawei Zhou
URL: https://arxiv.org/abs/2606.28186
Abstract:

Predicting human item difficulty is central to educational assessment, where reliable estimates support fairness and effective test construction. Existing methods often depend on costly human calibration or item-level textual representations, providing limited evidence about the cognitive processes that make items difficult. We argue that difficulty should be viewed not only as a property of item text, but also as an observable consequence of the problem-solving burden an item induces. Large Reasoning Models (LRMs) offer scalable process evidence through reasoning traces, but such evidence must be structured to support interpretable modeling. To this end, we introduce Epi2Diff (Episode to Difficulty), a framework that maps LRM reasoning traces into cognitively grounded episode sequences. These episodes group trace segments into functional problem-solving states, enabling difficulty to be modeled through reasoning scale, effort allocation, and state transitions. Epi2Diff extracts compact episode-dynamic features and combines them with semantic item representations for human difficulty prediction. Experiments on four real-world human difficulty datasets show that Epi2Diff consistently outperforms strong baselines, including fine-tuned small language models, LLM in-context learning, and supervised LLM adaptation. On SAT-derived classification benchmarks, Epi2Diff achieves an 8.1% average relative gain over supervised LLM fine-tuning baselines. Further analyses show that harder items induce more effortful, iterative, and implementation-centered episode dynamics, rather than merely longer responses. These results demonstrate that cognitive episodes in LRM reasoning traces provide a predictive and interpretable process representation for human item difficulty, offering a new lens for educational measurement with reasoning models.

13. LLawCo: Learning Laws of Cooperation for Modeling Embodied Multi-Agent Behavior

Authors: Qinhong Zhou , Chuang Gan , Anoop Cherian
URL: https://arxiv.org/abs/2606.28182
Abstract:

Embodied agents operating in decentralized and partially observable environments have attracted growing attention in recent years. However, existing large language model (LLM)-based agents often exhibit behaviors that are misaligned with their partners or inconsistent with the environment state, leading to inefficient cooperation and poor task success. To address this challenge, we propose a novel framework, Learning Laws of Cooperation (LLawCo), that enables embodied agents to autonomously align with both their partners and task objectives. Our framework allows agents to reflect on past failures to extract misaligned behavioral patterns, which are used to derive high-level behavioral laws, such as “Talk when necessary” and “Wait for partner.” These laws are explicitly incorporated into the agents’ chains of thought via supervised fine-tuning, aligning their reasoning with task requirements and the behavior of other agents. To evaluate our approach, we introduce PARTNR-Dialog, a large-scale multi-agent communicative and cooperative planning benchmark built on the PARTNR environment. Experiments on existing tasks and our new benchmark demonstrate significant improvements in cooperative efficiency and task success rates. Across four backbone LLMs, our method achieves average success rate improvements of 4.5% on the PARTNR-Dialog benchmark and 6.8% on the TDW-MAT benchmark over state-of-the-art open-source communicative agent frameworks. See the LLawCo project page for details: this https URL

14. Robust Harmful Features Under Jailbreak Attacks: Mechanistic Evidence from Attention Head Specialization in Large Language Models

Authors: Yanchen Yin , Dongqi Han , Linghui Li
URL: https://arxiv.org/abs/2606.28153
Abstract:

Jailbreak attacks bypass LLM safety alignment, yet their mechanisms remain poorly understood. We provide evidence that attacks do not comprehensively eliminate safety features, but instead selectively suppress specific attention heads. We identify two functionally differentiated types: Adversarially Compromised Heads (ACHs) concentrated in early layers, which are suppressed under attacks, and Safety-Aligned Heads (SAHs) in mid-layers, which maintain robust activations even when attacks succeed. Ablation studies support the causal role of ACHs and the contribution of SAHs to robust activations: suppressing a small number of ACHs is sufficient to induce jailbreak-like behavior on normally refused inputs, while removing SAHs substantially weakens mid-layer safety activations. Token-level attribution further shows that ACH suppression is driven specifically by attack-template tokens, providing a mechanistic account of why attacks can bypass refusal decisions through ACH suppression while leaving internal safety signals sustained by SAHs – a phenomenon we term Robust Harmful Features. To validate the practical significance of this robustness, we show that simply reading these persistent activations – without any training – yields competitive aggregate detection performance with strong adversarial robustness.

15. From Tokens to States: LLMs as a Special Case of World Models and the Continuous Path Beyond

Authors: Paul Dubois
URL: https://arxiv.org/abs/2606.28127
Abstract:

The AI community has framed the relationship between large language models (LLMs) and world models as a dichotomy: LLMs predict tokens; world models simulate reality. Yann LeCun argues in 2022 that reaching general intelligence requires abandoning autoregressive token prediction in favour of latent-space architectures. This framing is unnecessarily binary. Two claims will be defended. First, LLMs are a degenerate special case of world models: the state space is the set of all token sequences, the only action is appending one token, and world models are therefore a strict generalisation of LLMs, not a replacement. Second, there is a natural continuous spectrum from NTP to JEPA, with multi-token prediction, future-summary prediction, and next-latent prediction as intermediate stations already populated by current research. Moving along this spectrum relaxes the LLM constraints one by one. It also progressively surrenders the two practical advantages that make LLMs trainable at scale: internet-scale self-supervised data, and a transformer architecture co-designed for discrete token prediction. Both are examined as open research questions: the data question (the cliff from self-supervised text to instrumented action-labelled environments) and the architecture question (whether the transformer generalises to continuous-state prediction, or whether a new primitive is needed).

16. Single and Multi Truth Data Fusion using Large Language Models

Authors: Hira Beril Kucuk , Norman W Paton , Jiaoyan Chen , Zhenyu Wu
URL: https://arxiv.org/abs/2606.28062
Abstract:

Data fusion, also known as truth discovery, is a data integration problem that aims to determine the correct value or set of values for each attribute of an object when presented with potentially conflicting values from multiple sources. Data fusion tasks belong to two main categories: single-truth scenarios, where each attribute has only one correct value, and multi-truth scenarios, where multiple values can be valid simultaneously. This paper investigates the use of Large Language Models (LLMs) in data fusion tasks for tabular data. Various prompting strategies, encompassing both single-truth and multi-truth scenarios, are investigated empirically. Domain-dependent, domain-independent, zero-shot and one-shot prompts are evaluated on three different benchmark datasets. Experimental results demonstrate that LLM-based approaches outperform traditional unsupervised truth discovery methods, such as DART and LTM, across all datasets. The codebase of this study has been made publicly available on GitHub.

17. ToolPrivacyBench: Benchmarking Purpose-Bound Privacy in Tool-Using LLM Agents

Authors: Shijing Hu , Liang Liu , Zhu Meng , Zhicheng Zhao
URL: https://arxiv.org/abs/2606.28061
Abstract:

Large language models (LLMs) have increasingly moved from standalone text generation systems to agents that invoke external tools, access environments, and execute multi-step tasks. However, conventional function-calling benchmarks mainly evaluate task completion and API correctness, while privacy evaluation benchmarks typically focus on final responses or privacy judgments. Neither perspective captures purpose-bound information flow across an executed multi-tool trajectory. Motivated by this limitation in current agent evaluation, ToolPrivacyBench audits whether task-private atoms are routed only to authorized tools and downstream sinks, thereby evaluating both task completion and privacy over-disclosure during tool use. The benchmark contains 2,150 cases, including 1,150 fully synthetic privacy-sensitive business workflows and 1,000 cases adapted from existing multi-tool and function-calling benchmarks. Each case is represented by a policy knowledge base. After an agent executes against mock business backends, the evaluator compares recorded tool arguments and backend audit logs with this policy knowledge base. The evaluation covers nine widely used agents to characterize purpose-bound privacy over-disclosure. The results show that successful tool execution does not imply appropriate privacy disclosure: an agent may complete a task while transmitting unnecessary private information through intermediate tool calls. ToolPrivacyBench therefore formalizes a need-to-know disclosure boundary, under which each tool should receive only the information necessary for its stated purpose, and uses trajectory-level auditing to identify privacy over-disclosure in multi-tool workflows.

18. MultiHashFormer: Hash-based Generative Language Models

Authors: Huiyin Xue , Atsuki Yamaguchi , Nikolaos Aletras
URL: https://arxiv.org/abs/2606.28057
Abstract:

Language models (LMs) represent tokens using embedding matrices that scale linearly with the vocabulary size. To constrain the parameter footprint, prior work proposes hashing many tokens into a single vector within encoder-only models. While this offers parameter efficiency, many-to-one collisions prevent its use in causal LMs. In this paper, we propose MultiHashFormer, a new framework that allows hash-based autoregression. Each token is represented as a unique hash signature, a short sequence of discrete hash IDs, generated by multiple independent hash functions. A Hash Encoder compresses this signature into a single latent vector for processing by a Transformer decoder. Then, a Hash Decoder generates the hash signature of the next token, which is then mapped back to text. We evaluate our approach at the 100M, 1B and 3B parameter scales, demonstrating that MultiHashFormer consistently outperforms standard Transformer LMs across multiple benchmarks. Furthermore, we show that our model handles multilingual vocabulary expansion with a constant parameter footprint without any modifications.

19. Can LLMs Judge Better Than They Generate? Evaluating Task Asymmetry, Mechanistic Interpretability and Transferability for In-Context QA

Authors: Sambaran Bandyopadhyay
URL: https://arxiv.org/abs/2606.28050
Abstract:

LLM-as-a-Judge and self-evaluation pipelines implicitly assume that evaluation is easier than generation. We test this in a controlled in-context QA setting where a context passage is the sole information source and each model judges the answer it generated, removing the parametric-knowledge confound of open-domain comparisons. Across four benchmarks (SQuAD 2.0, DROP, HotpotQA, MuSiQue) and two models, evaluation is not uniformly easier: generation accuracy exceeds self-evaluation on three of four, with multi-hop MuSiQue the exception. Attention analysis reveals why: evaluation attends to context 3–5x less than generation does and barely reads the candidate answer. LoRA fine-tuning confirms the asymmetry is not a training artifact: generation fine-tuning induces over-acceptance and evaluation fine-tuning degrades generation. These findings challenge core assumptions in self-evaluation pipelines.

20. Dialogue to Detection: A Multimodal Hybrid NLP Pipeline for Insurance Fraud Detection

Authors: Muhammad Shakeel Akram , Amal Htait , Abdul Hamid Sadka , Emma Meisingseth , Karishma Jaitly
URL: https://arxiv.org/abs/2606.28002
Abstract:

Insurance fraud imposes substantial financial losses and operational inefficiencies, raising premiums and impacting trust among legitimate policyholders. Early detection at FNOL remains a persistent challenge. Existing approaches rely largely on private, text-only datasets, limiting progress on multimodal methods that integrate linguistic, behavioural, and speaker-based indicators. We introduce a synthetic multimodal framework that replicates FNOL conditions. It generates agent-customer dialogue transcripts and two-speaker audios, performs ASR and diarisation. Downstream modules combine NER, regex-based feature extraction, LLM-RAG retrieval, and speaker embeddings in a rule-based risk score to flag narrative reuse, structural inconsistencies, and cross-case voice repetition while balancing sensitivity and false positives. Dataset validation and component-level evaluations show stability and transfer potential, offering a reproducible baseline beyond text-only fraud detection.

21. From Black-Box to Clinical Insight: A Multi-Stage Explainable Framework for Speech-Based Cognitive Impairment Detection

Authors: Yasaman Haghbin , Sina Rashidi , Ali Zolnour , Fatemeh Taherinezhad , Ali Fartoot , Hossein Azadmaleki , James M Noble , Maryam Dadkhah , Maryam Zolnoori
URL: https://arxiv.org/abs/2606.27973
Abstract:

Speech-based cognitive impairment detection offers a noninvasive, accessible alternative to costly biomarker assays, yet transformer-based models remain clinically uninterpretable. We propose a multi-stage explainability framework that translates black-box transformer predictions into clinically grounded narratives by integrating SHapley Additive exPlanations (SHAP)-based token attribution, theory-informed linguistic features, and a four-stage LLM reasoning pipeline using LLaMA-3.1-70B-Instruct. Built on the SpeechCARE-Adaptive Gating Network multimodal screening model (F1 = 72.11% on the NIA PREPARE benchmark), the framework maps model outputs to four cognitive-linguistic dimensions, including lexical richness, syntactic complexity, and semantic coherence. Physician evaluation on 70 stratified English samples demonstrated strong alignment with patient-level cognitive profiles, and a System Usability Scale score of 82/100 indicated high potential for clinical workflow integration.

22. Two-Stage Fine-Tuning for Protein Sequence Generation with Targeted Amino-Acid Composition

Authors: Violeta Basten-Romero , Rubén Muñoz-Tafalla , Anna María Díaz-Rovira , Bertran Miquel-Oliver , Isaac Filella-Merce , Víctor Guallar
URL: https://arxiv.org/abs/2606.27939
Abstract:

Protein language models are standard priors for biological sequence generation, but steering them toward explicit distributional design targets remains largely unexplored. We study a constrained protein generation problem in which sequences must match a desired amino-acid (AA) composition profile while preserving plausible sequence statistics and diversity. The motivating application is synthetic feed protein design, where the AA composition of dietary proteins directly determines their nutritional value. We propose a two-stage pipeline in which domain-adaptive fine-tuning (FT) on an in-domain protein dataset is followed by iterative reward-weighted FT via reinforcement learning (RL) anchored against the FT model as a frozen reference. We evaluate the pipeline on two AA compositions and find that FT brings the average composition close to the target, while the subsequent RL enforces specific sequence constraints that FT alone cannot satisfy. We additionally evaluate the design choices of the proposed composition reward term against two baselines and an ablated variant, isolate the contribution of each training stage, and verify that AA composition alignment is achieved without degrading sequence quality.

23. Agentic AI-Powered Re-Identification: An Emerging, Scalable Threat to Mobility Microdata Privacy

Authors: Oscar Thees , Roman Müller , Matthias Templ
URL: https://arxiv.org/abs/2606.27936
Abstract:

The widespread collection of fine-grained location data by commercial data brokers creates a re-identification risk that is not widely recognised by the public. While prior research has established that mobility traces are highly unique and that individuals can, in principle, be identified from a handful of spatio-temporal points, such attacks have historically required significant manual effort from skilled analysts, limiting their practical scale. In this feasibility study, we demonstrate in a real world setting that agentic AI fundamentally changes this threat model. We present an end-to-end pipeline in which large language model agents autonomously search the open web, cross-reference public records and social media, and resolve raw coordinate sequences to candidate identities - without human intervention. We evaluate the pipeline on a spatio-temporal dataset containing simulated location points anchored at and around true home and work addresses, focusing on a high-risk disclosure scenario. Our results demonstrate that, from spatio-temporal data and public sources alone, our agentic AI successfully re-identified 18 of the 25 re-identifiable individuals (72%) and 18 of 43 cases overall (41.9%). We discuss implications for Statistical Disclosure Control (SDC) practice and outline the near-future escalation that data custodians and regulators must anticipate. De facto anonymity - an implicit foundation of SDC practice - is shifting. Agentic AI strengthens the case that re-identification is reasonably likely by any means under the GDPR Recital-26 standard, at costs of minutes-and-dollars per target.

24. Triadic Werewolf: A Jester Role for Multi-Hop Theory of Mind in LLMs

Authors: Avni Mittal
URL: https://arxiv.org/abs/2606.27909
Abstract:

Theory-of-mind evaluations of large language models typically use dyadic social-deduction games, where every observable cue points to a single hidden side, so a model with strong language priors can score well without ever simulating opponents’ incentives. We extend the Werewolf game with a Jester, a third faction whose utility on peer suspicion is inverted because it wins by being voted out, so optimal play requires reasoning across three opposing utility functions. Across 60 games on GPT-4.1, DeepSeek-V3.1, and Llama-3.3-70B with Jester self-learning on and off, the Jester wins 60-70% of games while Werewolves never exceed 20%, and GPT-4.1 wolves vote the Jester out on day 1 in 60-70% of games, a strictly self-defeating action. Self-learning helps DeepSeek and Llama but hurts GPT-4.1, with the cost landing on Villagers rather than Werewolves. Only DeepSeek learns the subtle strategy of looking suspicious without looking intentionally suspicious, and it gains the most from the loop. Triadic incentive structure exposes a layer of multi-agent reasoning that dyadic deduction games leave invisible.

25. A Study of Temporal Fusion Strategies for Named Entity Recognition in Historical Texts

Authors: Emanuela Boros
URL: https://arxiv.org/abs/2606.27881
Abstract:

Temporal variation poses a unique challenge for named entity recognition (NER) in historical texts, where entities drift in surface form and salience across time. While language models (LMs) have made progress in various NLP tasks, their ability to reason about temporality, especially in diachronic contexts, remains limited or at least, questionable. In this paper, we systematically study how temporal metadata can be structurally embedded into NER models using a range of lightweight fusion strategies. We experiment with both absolute and relative temporal representations, injected into Transformer-based architectures via early or late fusion mechanisms such as cross-attention, adapters, and concatenation. Our evaluations on French and German historical datasets reveal that late fusion strategies yield more robust and temporally generalisable performance, particularly in early and noisy periods.

26. SpatialUAV: Benchmarking Spatial Intelligence for Low-Altitude UAV Perception, Collaboration, and Motion

Authors: Haoyu Zhang , Meng Liu , Qianlong Xiang , Kun Wang , Yaowei Wang , Liqiang Nie
URL: https://arxiv.org/abs/2606.27876
Abstract:

Spatial intelligence is essential for low-altitude unmanned aerial vehicle (UAV) perception, collaboration, and navigation. However, existing UAV benchmarks often emphasize image-level recognition, single-view understanding, or narrow answer formats, leaving 3D spatial inference, multi-view collaboration, scene dynamics, and diverse task formulations insufficiently evaluated. To address these gaps, we introduce SpatialUAV, a real low-altitude UAV benchmark comprising 4,331 curated instances across 14 fine-grained task types, covering semantic discrimination, spatial relation, aerial–aerial collaboration, aerial–ground collaboration, and motion understanding. SpatialUAV organizes all samples into a unified visual-input–question–answer schema, while supporting seven input configurations and nine answer formats, including option labels, region identifiers, geometric values, cross-view correspondences, and free-form motion descriptions. To ensure reliable and grounded evaluation, our data construction pipeline integrates detector-assisted regions, depth supervision, metadata-derived rules, extensive manual annotation, blind filtering, and multi-turn human validation, together with task-specific metrics for heterogeneous outputs. Evaluating representative vision-language models across three categories, we show that current models remain far from human-level performance, with pronounced bottlenecks in cross-view association, structured grounding, geometric reasoning, and temporal viewpoint understanding. These results offer empirical guidance for advancing low-altitude UAV spatial intelligence. Code and data are available at this https URL .

27. Position Bias Correction is Insufficient for One-Pass Attention Sorting

Authors: Qiong Tang , Xiangkun Hu , Xiangyang Liu , Yiran Chen , Yunfan Shao
URL: https://arxiv.org/abs/2606.27793
Abstract:

Long-context language models suffer from position bias, where information in middle positions is underutilized. Attention Sorting addresses this by iteratively reordering documents based on attention patterns, but its multiple sort-and-generate cycles increase deployment cost. We hypothesize that position bias is the primary bottleneck and propose Debiased One-Pass Attention Sorting, which estimates a per-prompt position-bias curve from the low-attention majority of documents and uses it to correct raw attention scores (via subtraction or division) to enable single-pass sorting. Our experiments on two models refute this hypothesis in the tested setting: on LLaMA-2-7B-32K-Instruct, debiasing produces identical results to uncalibrated single-pass sorting (94.83\% containment accuracy), while on YaRN-Llama-2-7b-64k, debiasing improves accuracy by 8.67 percentage points but remains 14.84pp behind iterative sorting, closing only 37\% of the gap. These results suggest that position-bias correction is insufficient to match iterative sorting, and that repeated reordering provides additional benefits beyond bias correction.

28. NLL-Guided Full-Attention Layer Selection for Training-Free Sliding-Window Adaptation

Authors: Qiong Tang , Xiangkun Hu , Xiangyang Liu , Yiran Chen , Yunfan Shao
URL: https://arxiv.org/abs/2606.27791
Abstract:

Hybrid attention models that mix full and sliding-window attention across layers offer a promising approach to efficient long-context inference, but the critical question of \emph{which layers} should retain full attention remains unsolved. Existing methods use either fixed periodic patterns or attention-based heuristics that may not capture what matters for downstream accuracy. We propose NLL-guided layer selection, a training-free method that directly measures each layer’s importance by computing the negative log-likelihood degradation on answer tokens when that layer uses sliding-window instead of full attention. On LongMemEval with Qwen3-4B, our method achieves 64.6\% accuracy using only 1/4 full-attention layers, matching the 1/2-FA periodic baseline (65.0\%) while halving the computational budget. NLL-guided selection outperforms the SWAA-reported periodic 1/4-FA baseline by 10.4 percentage points and a matched LightTransfer-style baseline by 26.4 percentage points. De-confounding analysis shows the signal is consistent with long-range attention needs rather than generic layer sensitivity. The method requires only $\sim$15 minutes of one-time calibration, advancing the efficiency-accuracy Pareto frontier for long-context LLM deployment.

29. Output-Space Allocation Costs for Calibration-Guided LLM Compression: An Empirical Study

Authors: Qiong Tang , Xiangkun Hu , Xiangyang Liu , Yiran Chen , Yunfan Shao
URL: https://arxiv.org/abs/2606.27785
Abstract:

Training-free compression methods for large language models (LLMs) often use calibration data to guide compression decisions. ROCKET, a recent method combining sparse-dictionary factorization with multi-choice knapsack problem (MCKP) allocation, derives its per-layer factorization from an output reconstruction objective but uses weight-space Frobenius error as the MCKP allocation cost. We investigate whether aligning the allocation cost with the output-space objective improves compressed model fidelity. On Qwen3-8B at 50\% compression, our ROCKET-ActCost achieves +0.8 percentage points higher average accuracy across 8 zero-shot benchmarks (53.1\% vs 52.3\%), but increases WikiText perplexity by 16\% (61.46 vs 52.98). This accuracy-perplexity tradeoff reveals that different allocation objectives favor different downstream metrics. The high correlation ($>$0.99) between weight-space and output-space errors limits allocation divergence, explaining the modest effect size. On Llama-3.2-1B at 20\% compression, the two methods produce near-identical results (53.3\% vs 53.5\% accuracy, 14.45 vs 14.66 PPL), suggesting that the effect of the cost function is minor at lower compression ratios.

30. Drop-Then-Recovery: How Redundant Are Vision-Language-Action Models?

Authors: Guoheng Sun , Kaixi Feng , Shwai He , Xiaochuan Gong , Yexiao He , Ziyao Wang , Zheyu Shen , Wanghao Ye , Ramana Rao Kompella , Gaowen Liu , Ang Li
URL: https://arxiv.org/abs/2606.27755
Abstract:

Vision-Language-Action (VLA) models enable instruction-driven robotic manipulation, but they inherit oversized language backbones from pretrained VLMs whose capacity far exceeds what is needed for short robotic instructions. This raises a basic question: how much of a VLA model is actually necessary for closed-loop control? In this work, we study architectural redundancy in VLA models by using transformer block removal as a controlled intervention. We introduce \textbf{Drop-Then-Recovery (DTR)}, an analysis protocol that removes selected blocks from a pretrained VLA model and then fine-tunes the resulting model to measure whether the removed capacity was necessary for downstream control. To make this intervention reliable, we propose \textbf{GateProbe}, a one-shot virtual-gate sensitivity metric that ranks blocks by their contribution to the downstream action loss. Across multiple VLA architectures, manipulation benchmarks and even real-robot industrial scenarios, we find a strong asymmetry in post-removal recoverability: \ul{\textit{language backbones are highly redundant for standard robotic manipulation tasks, whereas vision and action pathways are substantially less tolerant to removal} }. On LIBERO, removing half of the LLM blocks even improves OpenVLA-OFT from 95.0% to 98.3% under the same downstream fine-tuning budget, and retaining only two language blocks still recovers baseline-level performance. These results suggest that current VLA benchmarks may exert limited pressure on deep language grounding and compositional instruction understanding, and that future VLA architectures should allocate capacity more deliberately across language, vision, and action components. The code is available at this https URL .

31. End-to-End Dynamic Sparsity for Resource-Adaptive LLM Inference

Authors: Yuhang Chen , Jinhao Duan , Ruichen Zhang , Mingfu Liang , Xiaohan Wei , Yunchen Pu , Fei Tian , Chonglin Sun , Parish Aggarwal , Frank Shyu , Luke Simon , Sandeep Pandey , Tianlong Chen , Xi Liu
URL: https://arxiv.org/abs/2606.27743
Abstract:

Large Language Models (LLMs) inference is typically deployed under a static resource assumption, where models execute a fixed computational graph regardless of the runtime environment. However, real-world cloud infrastructure is inherently dynamic, characterized by fluctuating availability (e.g., spot instance preemption) and tiered Quality-of-Service requirements. In such volatile settings, static models are inflexible: they either crash under resource constraints or waste compute on redundant operations. To bridge this gap, we propose Learning to Allocate (L2A), an end-to-end framework for resource-adaptive inference. Unlike prior methods that condition only on input difficulty, we formulate inference as a constrained allocation problem conditioned on both the input and the runtime resource budget itself. We introduce lightweight, budget-conditioned and input-aware gating networks integrated into the LLM. These gates are trained via a unified objective that jointly optimizes task performance, logical consistency, and resource costs along three axes matching how real-world dynamics manifest: layer skipping for memory and depth pressure, head pruning for throughput contention, and reasoning-token reduction for latency tightening. This lets the model learn a budget-aware policy beyond input difficulty alone: it adaptively configures its computational footprint with respect to real-time resource dynamics, maximizing reasoning depth when resources permit while enforcing strict frugality when budgets tighten. A single L2A model traces the entire compute-accuracy Pareto frontier on Llama-3-8B and Qwen-3-4B: at up to 34% realized layer sparsity, it stays within 0.6% of the dense baseline on GSM8K, with the same gap holding zero-shot on out-of-distribution tasks, while every static or heuristic baseline requires a separately tuned model and still drops by 5-10% at comparable inference time.

32. KG2Cypher: Data-Centric Pipeline for Building Enterprise Text-to-Cypher Systems

Authors: Minjun Choi , Yerin Kim , Junghyuk Seo , Sujin Mo , Hyemin Lee , Youngjoong Ko
URL: https://arxiv.org/abs/2606.27742
Abstract:

Enterprise Knowledge Graphs (KGs) are increasingly used for internal search, analytics, and question answering, but building natural-language interfaces for private enterprise graphs remains costly. We present KG2Cypher, a data-centric pipeline for building enterprise text-to-Cypher systems from existing KGs. KG2Cypher first constructs an executable Cypher query from observed graph facts and then uses LLMs to generate its associated natural-language question. The resulting Text-Cypher pairs are validated with an LLM judge and human validation, and are converted into candidate-aware SFT data. The trained generator is served with class-conditioned schema prompting, entity retrieval, and LoRA-based inference. We evaluate KG2Cypher in Korean enterprise settings, where short search-style queries and schema paraphrases make language grounding difficult. LoRA SFT improves execution-result F1 from 0.806 to 0.950 on broadcast-program queries and from 0.70 to 0.92 on company queries. In an 11-class setting, KG2Cypher achieves 95.2% exact match, 99.9% execution rate, and 0.964 execution-result F1.

33. Bifocal Diffusion Language Models: Asymmetric Bidirectional Context for Parallel Generation

Authors: Yuhang Chen , Xianfeng Wu , Jinhao Duan , Mingfu Liang , Xiaohan Wei , Yunchen Pu , Fei Tian , Chonglin Sun , Parish Aggarwal , Frank Shyu , Luke Simon , Sandeep Pandey , Xi Liu , Tianlong Chen
URL: https://arxiv.org/abs/2606.27732
Abstract:

Discrete diffusion language models (dLLMs) recover masked tokens in parallel, offering significant speedups over autoregressive (AR) generation. However, such promising frameworks face a fundamental architectural design dilemma: \ding{182} Adopting bidirectional attention achieves strong generation quality by allowing each position to access the full context, but is inherently incompatible with KV caching, limiting inference throughput in batch-serving scenarios; \ding{183} Conversely, causal attention enables efficient cached inference but loses all right-side context, substantially degrading generation quality. This paper introduces Bifocal dLLMs, a new paradigm that resolves this dilemma through \emph{asymmetric bidirectional context}. Analogous to bifocal lenses, we instantiate the paradigm as \textbf{R2LM} (Right-to-Left Mamba), which combines two complementary mechanisms: $a$) standard causal attention providing precise left-context with full KV cache compatibility, while $b$) a lightweight reverse Mamba SSM sidecar supplying compressed right-side context without breaking cacheability. Comprehensive experiments on continued pretraining of Qwen3-1.7B with 60B tokens demonstrate that R2LM achieves $2.4\times$ to $12.9\times$ higher throughput than bidirectional dLLMs and $1.9\times$ to $2.9\times$ speedup over AR baselines in batch serving through parallel decoding with KV caching, while exceeding the causal baseline on most benchmarks and surpassing the bidirectional dLLM on average.

34. Enhancing Numerical Prediction in LLMs via Smooth MMD Alignment

Authors: Zhuo Zuo , Li Yue , Wenhao Zheng , Chenpeng Wang , Xianggen Liu
URL: https://arxiv.org/abs/2606.27731
Abstract:

Despite their strong general capabilities, large language models (LLMs) often remain unreliable when outputs must be numerically precise. A key reason is the training objective: standard cross-entropy treats numeric tokens as unstructured categories and ignores the metric structure of their values. We address this mismatch with Smooth Maximum Mean Discrepancy (SMMD), which builds on the classic MMD by incorporating value-distance kernels over numeric tokens and graph-based smoothness. With this kernel defined over a numeric sub-vocabulary, SMMD aligns the predicted numeric distribution to the target via kernel matching and smooths the prediction-target residual over the induced kernel graph to encourage local consistency. We evaluate SMMD on four numeric-target tasks: mathematical reasoning, arithmetic calculation, clock-time recognition, and chart question answering, across multiple open-weight LLM and VLM backbones. SMMD consistently improves accuracy over both cross-entropy and recent numeric-target losses; analyses show complementary effects between MMD and smoothness and underscore the importance of distance-based kernel design. Code is available at this https URL .

35. Low-Agreeableness Persona Conditioning for Safe LLM Fine-Tuning

Authors: Austin MY Cheung , Yi Yang
URL: https://arxiv.org/abs/2606.27709
Abstract:

Recent work has shown that fine-tuning large language models (LLMs) for social warmth degrades factual reliability and increases sycophancy. We investigate a related but distinct failure mode: warmth fine-tuning also weakens adversarial safety, making models more susceptible to jailbreaks and harmful output generation. We examine whether this reflects an inherent consequence of empathetic adaptation or an artifact of data construction. To address this, we introduce a persona-driven rewriting pipeline that conditions user turns on low agreeableness and pairs this with warm, de-escalating assistant responses. Across three experiments on four models, our approach reduces jailbreak susceptibility and harmful output rates relative to generic warmth fine-tuning baselines, while preserving conversational warmth. Representational probing provides suggestive evidence that this conditioning reduces the geometric alignment between warmth and compliance directions in latent space. These results show that safer empathetic fine-tuning is achievable through data design alone, without safety labels, harm detectors, or changes to the training objective.

36. Mitigating LLM-based p-Hacking by Preregistering for the Next LLM

Authors: Maria Thomas , Kristina Gligoric , Nihar B. Shah
URL: https://arxiv.org/abs/2606.27687
Abstract:

Large language models (LLMs) are increasingly used to generate, classify, and annotate data whose outputs feed downstream hypothesis tests. However, LLM-based research is easy to p-hack: a researcher can tune the prompts, decoding parameters, or output format until a desired result is reached. We propose a protocol to mitigate p-hacking in LLM-based research: preregistering the experiment and eligible models, and then running it on the first eligible LLM that is released after the preregistration. The researcher finalizes the procedure on current models, preregisters the analysis plan together with a set of eligible future models, and runs the confirmatory analysis on the first eligible model released afterward. Because this model does not exist at commitment time, it cannot be hacked against; furthermore, configurations that hack one model frequently do not transfer to the next. We evaluate the protocol on two tasks whose true values are known. Across 20 models from four providers and 11 LLM-analysis configurations, the protocol would have blocked successful transfer of the p-hack in 73.9% and 72.7% of cases in the two tasks. Additional analyses reveal that mitigation remains substantial under several stress tests. Finally, putting money where our mouth is, we followed our own protocol and preregistered our experiment. The preregistered experiment confirmed the protocol’s effectiveness: out of the 7 configurations that hacked the prior model, the hacking failed to carry over in 6 configurations on the first eligible model released afterward.

37. CBD: API-Only LLM Black-Box Unlearning through Controlled Behavioral Divergence

Authors: Zhiqiang Xie , Yijing Lin , Zhipeng Gao , Dong In Kim
URL: https://arxiv.org/abs/2606.27683
Abstract:

Edge devices increasingly invoke large language models (LLMs) through API services for context aware edge intelligence, while edge generated data may be collected to improve LLMs and may introduce sensitive, copyrighted, harmful, or outdated information into model behavior. Machine unlearning offers a practical way to remove the influence of undesired data without retraining LLMs. However, existing methods still face two gaps. The first is API only black box access, where target model parameters and internal logits are unavailable. The second is how to preserve retained utility when unlearning target data and retained data share highly similar prompt structures or semantic patterns. To address these challenges, we propose Controlled Behavioral Divergence (CBD), an API only black box unlearning framework. CBD uses two auxiliary models to create controlled behavioral divergence between retained inputs and unlearning target inputs, converts this divergence into an unlearning relevance score, and routes unlearning related prompts away from the target LLM. To improve discrimination accuracy under high similarity between target and retained data, CBD constructs a gradient statistics based discriminative basis by estimating empirical Fisher matrices and solving a regularized generalized eigenvalue problem, guiding the unlearning signal toward target specific information rather than shared prompt structures. Compared with eleven white box and gray box unlearning baselines, CBD achieves a better unlearning utility trade off and its performance varies little across settings. On ToFU forget10, CBD approaches the retrained reference on the forget set while raising model utility to 74.90, about 15% above the second best baseline. On WMDP, it lowers hazardous knowledge accuracy to 25.68, near random guessing, while preserving MMLU accuracy of 52.67. Code is at this https URL .

38. From Signals to Transfer: A Factorised Study of Probe-Based Uncertainty Estimation in Large Language Models

Authors: Ponhvoan Srey , Xiaobao Wu , Cong-Duy Nguyen , Quang Minh Nguyen , Duc Anh Vu , Anh Tuan Luu
URL: https://arxiv.org/abs/2606.27679
Abstract:

Probe-based uncertainty estimation (UE) has emerged as a prominent approach to detect hallucinations in Large Language Models (LLMs) by learning uncertainty from internal model signals. Yet, recent methods vary simultaneously across feature design, training data construction, and evaluation setting, obscuring what actually drives performance. To address this issue, we propose a factorised study of probe-based UE under matched conditions. Our results show that raw hidden states and attention features are difficult to outperform in-domain. However, under distribution shift, structured and compressed features are more robust, suggesting that in-domain performance alone is insufficient to measure progress. Furthermore, prompting and label construction significantly affect probe behaviour. Building on these best-practice findings, we train benchmark-based pretrained probes that transfer reasonably well to open-ended factual generation, providing a stable off-the-shelf baseline. Our work encourages more deployment-oriented evaluation of probe-based uncertainty estimators. The code repository is available at this https URL .

39. HybridCodec: Modeling Discrete and Continuous Representations for Efficient Speech Language Models

Authors: Artem Ploujnikov , Francesco Verdini , Samir Sadok , Mirco Ravanelli
URL: https://arxiv.org/abs/2606.27627
Abstract:

Discrete audio representations have become increasingly popular for building multimodal text-audio systems and integrating audio capabilities into Large Language Models (LLMs). However, numerous studies report performance degradation on various downstream tasks due to information loss during discretization. To address this, we propose a novel approach combining temporally compressed discrete tokens with dimensionality-reduced continuous residuals. Our framework consists of a hybridized discrete-continuous focal modulation codec and a hybrid Transformer. This architecture performs autoregressive inference in the discrete domain, coupled with non-autoregressive prediction and continuous residual upsampling. Experimental results show that our approach significantly improves the retention of speaker characteristics compared to discrete-only methods, while simultaneously reducing the number of required autoregressive steps.

40. Dismantling Pathological Shortcuts: A Causal Framework for Faithful LVLM Decoding

Authors: Liu Yu , Can Chen , Ping Kuang , Zhikun Feng , Fan Zhou , Gillian Dobbie
URL: https://arxiv.org/abs/2606.27596
Abstract:

Large Vision-Language Models (LVLMs) exhibit sophisticated reasoning but remain susceptible to object hallucination. Deviating from the prevailing attention intensity assumption, we reveal a deeper dynamic structural misalignment: hallucination is triggered at decision-critical steps where specific attention heads, acting as risky mediators, decouple from visual evidence to lock onto language priors. This establishes a pathological shortcut that bypasses visual grounding. To dismantle this, we propose Fox (Faithfulness and Observational-flow via eXpression-rectification), a training-free inference-time framework. Fox diagnoses structural misalignment using a visual attention entropy probe to localize risky mediators unsupervisedly. We then execute a targeted causal intervention via numerical logit saturation to physically sever the shortcut path. Finally, a conflict-gated cooperative decoding strategy reconciles interventional faithfulness with observational fluency. Extensive experiments demonstrate that Fox achieves SOTA performance, outperforming SID by 29.1% while preserving linguistic richness. Code is available at this https URL .

41. On the Inseparability of Instructions and Data in Shared-Embedding Sequence Models

Authors: Dewank Pant , Shruti Lohani , Avijit Kumar
URL: https://arxiv.org/abs/2606.27567
Abstract:

Prompt injection is the top security risk for LLM-integrated applications, yet every defense proposed so far has been broken. We prove this is not a coincidence: in shared-embedding architectures that lack enforced control-data separation, perfect prompt-injection prevention is mathematically impossible. We formalize prompted systems as Prompted Action Models whose outputs include control-authoritative actions: refusal decisions, tool authorization, policy routing, and memory writes. We define Semantic-Faithful Control (SFC), the property that such behavior depends only on the meaning of untrusted input, not on how it is encoded. We then prove SFC is unachievable within the shared pipeline, via three results: a provenance-recovery impossibility (shared representations make trusted and untrusted content statistically inseparable, bounded by total variation distance); control-path exposure (untrusted tokens enter control-relevant computation through the same attention value-aggregation that determines outputs); and a finite-coverage invariance gap (finite training cannot certify invariance over infinite semantic-equivalence classes). We ground each quantity in measurements on production tokenizers and models. The result is structural, not a gap in current defenses. It mirrors the code-data confusion in Von Neumann machines that gives rise to buffer overflows, a vulnerability class that took decades of layered defenses (DEP, Write-XOR-Execute, ASLR, stack canaries, and ultimately memory-safe languages) to contain, because no single mechanism sufficed. The implication is the same: prompt injection cannot be eliminated by better in-pipeline classification or alignment alone. It requires architectural separation of instruction and data channels. We identify the root cause and the class of solution it demands.

Authors: Utkarsh Sahu , Zhisheng Qi , Li Zhu , Yizhao Yang , Jun Li , Ryan Rossi , Yu Wang
URL: https://arxiv.org/abs/2606.27539
Abstract:

Social media popularity prediction aims to forecast the future reach or influence of online content from early-stage observations. Accurate prediction enables key downstream applications, such as advertising optimization and strategic content planning by users, creators, and platforms. Despite substantial progress, existing popularity prediction works often fail to jointly consider multimodal content and temporal social interaction signals. Moreover, the literature remains highly fragmented across datasets, modalities, observation windows, prediction targets, and evaluation protocols. This fragmentation prevents fair comparison and obscures a systematic understanding of how textual, visual, temporal, and interaction-based signals jointly shape popularity dynamics. To address these challenges, we introduce MMG-Pop, a Multi-modal Graph-based Popularity Prediction benchmark, which unifies datasets, modalities, temporal interaction signals, and representative baselines under a standardized evaluation protocol. Furthermore, we propose MMG-PopNet, a unified multi-modal graph-based network that jointly models the aforementioned multi-modal signals and graph-structured social interactions. Extensive experiments on MMG-Pop, comprising four datasets across Bluesky and Reddit platforms, demonstrate the superior performance of MMG-PopNet and yield new insights into cross-platform training generalization, multi-task prediction benefits, multi-modality contributions, and LLM prediction limitation. These findings establish a unified foundation for future research on social dynamics modeling and intervention under heterogeneous modalities and socially-aware agentic ecosystem paradigms.

43. Large Language Model Teaches Visual Students: Cross-Modality Transfer of Fine-Grained Conceptual Knowledge

Authors: Thomas Shih-Chao Liang , Zhuoran Yu , Yong Jae Lee
URL: https://arxiv.org/abs/2606.27527
Abstract:

Large Language Models (LLMs) possess broad conceptual knowledge acquired through large-scale text pretraining, yet their potential to supervise models in other modalities remains underexplored. In this work, we propose LaViD–Language-to-Visual Knowledge Distillation–a simple and effective framework for transferring high-level semantic knowledge from a language-only teacher to a vision-only student model. Instead of relying on paired multimodal data, LaViD elicits conceptual signals from an LLM by prompting it to generate multiple-choice questions (MCQs) that probe semantic distinctions between visual classes. Each class is mapped to a soft label distribution over these MCQs, forming a rich conceptual signature that guides the student through an auxiliary distillation loss. Notably, despite using a language-only teacher without access to image data, LaViD consistently outperforms recent methods like MaKD that distill from vision-language models across multiple fine-grained benchmarks. It also achieves competitive or superior performance compared to state-of-the-art visual distillation methods such as DKD and MLKD, with further gains when combined with logit standardization. On the Waterbirds dataset, LaViD substantially improves worst-group accuracy, demonstrating enhanced robustness to spurious correlations with distillation. Code is available at this https URL .

Authors: Aditi Gupta , Neel Mishra , Kushagra Trivedi , Pawan Kumar
URL: https://arxiv.org/abs/2606.27474
Abstract:

How should we evaluate generation systems that combine autoregressive (AR) and diffusion decoding? We study this question through Speculative Refinement (SpecRef), a training-free hybrid method that warm-starts a masked diffusion language model from an AR draft using entropy-guided selective masking. Evaluating SpecRef across six benchmarks (HumanEval, MBPP, GSM8K, BBH, ARC-Challenge, HellaSwag) with three distinct evaluation protocols (execution-based pass@1, exact-match, log-likelihood scoring), we surface several findings relevant beyond our specific system: (1) code benchmarks conflate structural discovery with logical correctness: providing a syntactic scaffold lifts accuracy from near zero to over 20% without changing the model, indicating that much of the baseline failure is structural; (2) a refinement tension phenomenon where multi-stage correction degrades already-correct tokens, exposing benchmark saturation ceilings invisible to single-model evaluation; (3) log-likelihood and generative evaluation produce different model rankings for the same model pair, suggesting they measure different capabilities; (4) standard Python post-processing silently breaks code evaluation for non-AR generators. These observations apply to any multi-stage or non-autoregressive generation pipeline and point toward more diagnostic evaluation practices.

45. Supersede: Diagnosing and Training the Memory-Update Gap in LLM Agents

Authors: Vedant Patel
URL: https://arxiv.org/abs/2606.27472
Abstract:

Large language model (LLM) agents operate over long, multi-session interactions in which facts change: a user moves, a price updates, a plan is revised. Acting correctly requires using the current value of a fact and discarding values that have been superseded. We isolate this ability on real conversational data and show that it is a distinct, unsolved failure. On the knowledge-update subset of LongMemEval, replacing an agent’s full context with a bounded, self-maintained memory drops accuracy from 92% to 77% even on a frontier model (gpt-5.4), a gap that is statistically significant (paired McNemar p<0.005) and persists across model scale while full-context accuracy saturates near 92%. The bottleneck is therefore memory maintenance, not comprehension, and is not closed by a stronger model. We then ask whether this is merely an undersized memory, and find it is not: as the conversation grows 24x, accuracy falls further (from 68% to 28%), and granting the agent proportionally more memory yields no detectable recovery (28% to 28%, n=25). The failure scales with the length of the conversation, not the compression ratio. We release Supersede, an open reinforcement-learning environment (on the verifiers / prime-rl stack) that turns this measurement into a training signal: agents are rewarded for answering from the current value and penalized for stale ones. Finally, we close the loop and show the gap is trainable: GRPO fine-tuning a small open model (Qwen2.5-3B) on this environment nearly doubles its held-out supersession accuracy on real, unseen conversations (9.0% to 16.7%, a single run), along a monotonic checkpoint curve indicating the learned policy, not the harness, carries the gain. To our knowledge this is the first trainable environment whose reward targets temporal fact-currency, and the first evidence the supersession gap can be trained down, not only measured.

46. SidConArena: An Environment Evaluating Agents in Open-Ended,Positive-Sum Bargaining Game

Authors: Yeqi Feng , Yuxin Chen , Tianxing He
URL: https://arxiv.org/abs/2606.27397
Abstract:

Evaluating LLM agents requires dynamic environments that go beyond static reasoning and zero-sum games. Real-world economic interaction is often open-ended and mixed-motive: agents must negotiate, create positive-sum surplus, compete for scarce assets, and plan under delayed returns. We introduce SidConArena, a new benchmark framework for evaluating LLM agents in open-ended, positive-sum bargaining. SidConArena formalizes a multi-player economy as a finite-horizon partially observable stochastic game with three coupled phases: natural-language negotiation with binding trades, deterministic converter-based production, and sealed-bid auctions for long-term assets. The framework combines structured observations, phase-aware agent dispatching, a neural-symbolic action interface, and asynchronous execution, enabling free-form interaction while preserving rule-grounded evaluation. Across homogeneous and heterogeneous tournaments, stronger frontier models achieve higher economic outcomes, yet agents still misvalue resources, bargain passively, and remain limited in long-horizon investment planning.

47. Agentic Publication Protocol: An Attempt to Modernize Scientific Publication

Authors: Sirui Lu , Xiao-Liang Qi
URL: https://arxiv.org/abs/2606.27386
Abstract:

Scientific publication is still organized primarily around static manuscripts, even though much of scientific progress depends on tacit know-how: how to run code, reproduce figures, interpret edge cases, choose useful follow-up directions, and avoid failed paths. Large language model agents create an opportunity to publish not only knowledge, but also operational know-how in a form that future readers and researchers can directly use. This paper outlines the Agentic Publication Protocol (APP), a lightweight repository format for packaging a paper together with code, data, environment information, reproducibility instructions, and an agent-facing instruction file. APP treats a version-controlled repository as the publication object and uses \texttt{ this http URL } and optional skills to define a paper agent that can explain the work, reproduce key results when possible, and support follow-up research. We describe the design principles and details of the protocol, as well as the agent skills useful for publishing papers under the protocol. We also describe development tools for evaluating and improving the protocol and associated agent skills. Finally, we provide a broader discussion of the future of scientific research in the agent era.

48. CalBrief: A Pilot Diagnostic Benchmark for Evidence-Calibrated Scientific Briefing with Large Language Models

Authors: Yu Fu , Yongqi Kang , Yong Zhao
URL: https://arxiv.org/abs/2606.27383
Abstract:

Large language models (LLMs) are increasingly used as research assistants, yet it remains unclear whether they can calibrate research takeaways to the strength and scope of the supporting evidence. We study evidence-calibrated scientific briefing: given a bounded package of related papers, a system should generate package-level takeaways with evidence strength, scope boundaries, and missing-evidence caveats. We contribute a verified pilot benchmark of 16 heterogeneous scientific evidence packages and 96 human-verified takeaways, and we use CalBrief, an auditable role/gap/strength framework, as a diagnostic probe to locate where briefing breaks down. Under a fair-schema evaluation, structured organization improves role and gap reasoning, but an explicit strength-calibration policy is systematically over-conservative and falls below majority and direct-LLM baselines. To explain why, we run a controlled diagnostic across three closed-model backbones (GPT-4o, Claude Sonnet, Gemini Flash) that separates three potential causes of conservatism. Approximately 63% of the conservatism gap is attributable to expanding the label space from binary {moderate, weak} to four-way {moderate, weak, uncertain, insufficient_evidence} (p < 0.001 across all backbones); only 1% is attributable to gap/scope signal injection (not significant); the remaining 36% arises from the pipeline policy itself. We also find that 4-way predictions can be post-hoc collapsed back to binary and then match or exceed direct binary prompting, so the extra labels carry information that strict matching hides. Label-level strength judgment and auditable evidence organization are distinct abilities currently in tension, and should be evaluated separately for LLM research assistants.

49. Position: The Term “Machine Unlearning” Is Overused in LLMs

Authors: Sangyeon Yoon , Yeachan Jun , Albert No
URL: https://arxiv.org/abs/2606.27379
Abstract:

Large language models increasingly face demands to “forget” training data, knowledge, or behaviors due to regulatory deletion obligations, copyright/licensing disputes, and safety or product-policy requirements. This position paper argues that machine unlearning is overused as a term in LLM research and should be reserved for dataset-defined deletion: removing the training influence of a precisely specified forget set such that the resulting model is approximately indistinguishable from retraining without that data. We contend that many tasks currently labeled “unlearning” (e.g., refusal for harmful requests, entity/knowledge removal, or targeted suppression) pursue different, often policy-dependent objectives and therefore require different terminology and baselines (e.g., alignment, suppression, editing, obfuscation). We further argue that this confusion is not cosmetic: because papers make different implicit guarantees under the same label, metrics and benchmarks are frequently reused outside their intended scope, rewarding surface-level non-disclosure (e.g., low ROUGE/forget accuracy) even when retraining-equivalence is not tested and derived capabilities remain. We conclude by calling for stricter terminology tied to explicit guarantees and reference models, and for evaluations that match the claimed objective.

50. DataStates-LLM: Scalable Checkpointing for Transformer Models Using Composable State Providers

Authors: Avinash Maurya , M. Mustafa Rafique , Franck Cappello , Bogdan Nicolae
URL: https://arxiv.org/abs/2601.16956
Abstract:

The rapid growth of Large Transformer-based models, specifically Large Language Models (LLMs), now scaling to trillions of parameters, has necessitated training across thousands of GPUs using complex hybrid parallelism strategies (e.g., data, tensor, and pipeline parallelism). Checkpointing this massive, distributed state is critical for a wide range of use cases, such as resilience, suspend-resume, investigating undesirable training trajectories, and explaining model evolution. However, existing checkpointing solutions typically treat model state as opaque binary blobs, ignoring the 3D heterogeneity'' of the underlying data structures--varying by memory location (GPU vs. Host), number oflogical’’ objects sharded and split across multiple files, data types (tensors vs. Python objects), and their serialization requirements. This results in significant runtime overheads due to blocking device-to-host transfers, data-oblivious serialization, and storage I/O contention. In this paper, we introduce DataStates-LLM, a novel checkpointing architecture that leverages State Providers to decouple state abstraction from data movement. DataStates-LLM exploits the immutability of model parameters during the forward and backward passes to perform ``lazy’’, non-blocking asynchronous snapshots. By introducing State Providers, we efficiently coalesce fragmented, heterogeneous shards and overlap the serialization of metadata with bulk tensor I/O. We evaluate DataStates-LLM on models up to 70B parameters on 256 A100-40GB GPUs. Our results demonstrate that DataStates-LLM achieves up to 4$\times$ higher checkpointing throughput and reduces end-to-end training time by up to 2.2$\times$ compared to state-of-the-art solutions, effectively mitigating the serialization and heterogeneity bottlenecks in extreme-scale LLM training.

LLM 관련 주요 논문 - 2026-06-29

1. Tandem Reinforcement Learning with Verifiable Rewards

2. JD Oxygen AI Item Center (Oxygen AIIC) V1: An Industrial-Scale LLM/VLM-Centric Solution for Item Understanding, Management, and Applications

3. NormAct: A Benchmark for Hidden Social Norm Compliance in Embodied Planning

4. Grounded Iterative Language Planning: How Parameterized World Models Reduce Hallucination Propagation in LLM Agents

5. Towards Reliable and Robust LLM Planning: Symbolic Feedback-Driven Iterative Self-Refinement Framework

6. ToE: A Hierarchical and Explainable Claim Verification Framework with Dynamic Multi-source Evidence Retrieval and Aggregation

7. DysLexLens: A Low-Resource LLM Framework for Analysing Dyslexic Learners Insights from Online Forums

8. Odyssey: Constructing Verifiable Local Truth-Preserving Foundation Models

9. Internalizing the Future: A Unified Agentic Training Paradigm for World Model Planning

10. When Does Personality Composition Matter for Multi-Agent LLM Teams?

11. AI-Model Network: Concept, Current State and Future

12. Cognitive Episodes in LLM Reasoning Traces Enable Interpretable Human Item Difficulty Prediction

13. LLawCo: Learning Laws of Cooperation for Modeling Embodied Multi-Agent Behavior

14. Robust Harmful Features Under Jailbreak Attacks: Mechanistic Evidence from Attention Head Specialization in Large Language Models

15. From Tokens to States: LLMs as a Special Case of World Models and the Continuous Path Beyond

16. Single and Multi Truth Data Fusion using Large Language Models

17. ToolPrivacyBench: Benchmarking Purpose-Bound Privacy in Tool-Using LLM Agents

18. MultiHashFormer: Hash-based Generative Language Models

19. Can LLMs Judge Better Than They Generate? Evaluating Task Asymmetry, Mechanistic Interpretability and Transferability for In-Context QA

20. Dialogue to Detection: A Multimodal Hybrid NLP Pipeline for Insurance Fraud Detection

21. From Black-Box to Clinical Insight: A Multi-Stage Explainable Framework for Speech-Based Cognitive Impairment Detection

22. Two-Stage Fine-Tuning for Protein Sequence Generation with Targeted Amino-Acid Composition

23. Agentic AI-Powered Re-Identification: An Emerging, Scalable Threat to Mobility Microdata Privacy

24. Triadic Werewolf: A Jester Role for Multi-Hop Theory of Mind in LLMs

25. A Study of Temporal Fusion Strategies for Named Entity Recognition in Historical Texts

26. SpatialUAV: Benchmarking Spatial Intelligence for Low-Altitude UAV Perception, Collaboration, and Motion

27. Position Bias Correction is Insufficient for One-Pass Attention Sorting

28. NLL-Guided Full-Attention Layer Selection for Training-Free Sliding-Window Adaptation

29. Output-Space Allocation Costs for Calibration-Guided LLM Compression: An Empirical Study

30. Drop-Then-Recovery: How Redundant Are Vision-Language-Action Models?

31. End-to-End Dynamic Sparsity for Resource-Adaptive LLM Inference

32. KG2Cypher: Data-Centric Pipeline for Building Enterprise Text-to-Cypher Systems

33. Bifocal Diffusion Language Models: Asymmetric Bidirectional Context for Parallel Generation

34. Enhancing Numerical Prediction in LLMs via Smooth MMD Alignment

35. Low-Agreeableness Persona Conditioning for Safe LLM Fine-Tuning

36. Mitigating LLM-based p-Hacking by Preregistering for the Next LLM

37. CBD: API-Only LLM Black-Box Unlearning through Controlled Behavioral Divergence

38. From Signals to Transfer: A Factorised Study of Probe-Based Uncertainty Estimation in Large Language Models

39. HybridCodec: Modeling Discrete and Continuous Representations for Efficient Speech Language Models

40. Dismantling Pathological Shortcuts: A Causal Framework for Faithful LVLM Decoding

41. On the Inseparability of Instructions and Data in Shared-Embedding Sequence Models

42. Benchmarking Multi-Modal Graph-based Social Media Popularity Prediction

43. Large Language Model Teaches Visual Students: Cross-Modality Transfer of Fine-Grained Conceptual Knowledge

44. Speculative Refinement: A Hybrid Autoregressive Diffusion Decoding Strategy and Its Behavior Across Benchmarks

45. Supersede: Diagnosing and Training the Memory-Update Gap in LLM Agents

46. SidConArena: An Environment Evaluating Agents in Open-Ended,Positive-Sum Bargaining Game

47. Agentic Publication Protocol: An Attempt to Modernize Scientific Publication

48. CalBrief: A Pilot Diagnostic Benchmark for Evidence-Calibrated Scientific Briefing with Large Language Models

49. Position: The Term “Machine Unlearning” Is Overused in LLMs

50. DataStates-LLM: Scalable Checkpointing for Transformer Models Using Composable State Providers