Every paper checked — nothing hidden
All papers
Every paper checked in the week of 12–18 Oct 2026 · Multimodal models — nothing is hidden. Brief papers first.
All1,924In brief3More (on-topic)553Not relevant1,333Not assessed19Older33Could not confirm19Blogs with no paper11All blogs13
| Paper | Topic | Relevance | Feedback |
|---|---|---|---|
| Moral Combat: Exploiting Moral Foundations Theory to Jailbreak LLMs The paper introduces a multi-agent framework that reframes harmful requests using Moral Foundations Theory to test model jailbreak vulnerabilities. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Studies text prompt reframing attacks, without multimodal data. | |
| PA1-OPD: One-Step Prefix-Augmented On-Policy Distillation for Language Model Pretraining The paper proposes a prefix-augmented distillation method providing one-step student branch feedback during language model pretraining. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on text language model pretraining distillation. | |
| PN-Mix: Pair-Norm Guided Mixed-Precision Key Cache Quantization for Efficient Long-Output LLM Inference The paper proposes a pair-norm guided mixed-precision quantization method for key caches during long-output transformer inference. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Applies to decoder-only text language models without multimodal data. | |
| Twenty Days Later: A Pre-Registered Test of History-Based Forecasts of Brand Mentions in Gemini Answers The paper reports a pre-registered test measuring the stability of history-based forecasts for model brand mentions over twenty days. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Analyzes text query answers only, without multimodal components. | |
| J-tau: A Fully Localized Japanese Benchmark for Tool-Agent-User Interaction The paper introduces a localized Japanese benchmark for evaluating tool-using conversational agents in customer-support scenarios. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Evaluates text-based dialogue and tool interactions without multimodal capabilities. | |
| Selective Reasoning for Typed Decision Models: When to Start, Read, and Stop The paper proposes an interface for typed decision models that selectively invokes and interrupts reasoning to balance accuracy and latency. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Targets text reasoning models without multimodal inputs. | |
| Beyond Distribution Matching: Selective Margin Preservation for Supervised Fine-Tuning The paper introduces a soft margin preservation target for supervised fine-tuning to retain non-target token rankings. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses entirely on text-based language model supervised fine-tuning. | |
| Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models The paper proposes a training-free interface for heterogeneous language models to communicate facts using shared symbolic word indices. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on cross-model hidden state symbol communication in text models. | |
| BANGLAPLAN: A Benchmark for Constraint-Based Planning and Decision-Making in Bangla The paper introduces a 1,000-problem benchmark for constraint-based planning and decision-making in the Bangla language. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Text-only planning benchmark in Bangla without vision or audio modalities. | |
| RouteWeaver: Weaving Mode Selection and Execution into Unified LLM Routing The paper introduces a unified routing framework that learns both execution mode selection and model assignment across diverse tasks. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Addresses routing across language tasks without multimodal capabilities. | |
| Beyond State Capacity: Why Retrieval Breaks in Mamba The paper analyzes retrieval failures in Mamba architectures under query-first settings and explores token representations and selection gates. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Investigates text-based recurrent state mechanics, not multimodal reasoning. | |
| MakSafe: A Spatiotemporal Semantic Framework for Formal Specification and Compliance Checking of Road Vehicle Safety Rules The paper presents a formal spatiotemporal framework to specify and verify road vehicle safety rules across continuous trajectories. Peer-reviewed | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Formal spatiotemporal rule checking, not multimodal foundation models. | |
| Beyond Reachability: A Relation-Semantic Reasoning Framework for Biomedical Multi-hop QA The paper proposes a relation-semantic reasoning framework to resolve disconnects between answer reachability and valid paths in biomedical graphs. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Knowledge graphs and text, not multimodal models. | |
| Qwen-MT-Lumen at WMT2026: Instruction-Specialized Translation via Quality-Aware Training and On-Policy Distillation The paper presents a four-stage training pipeline combining preference optimization and distillation for instruction-following machine translation. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Text translation only, not multimodal. | |
| Likelihood Is Not Authorship: Can We Reliably Measure Human Contribution in AI-Assisted Writing? The paper reveals vulnerabilities in prompt-predictability metrics used to estimate human contribution in AI-assisted writing. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Text attribution, not multimodal models. | |
| More Methods, Better Agents? A Reality Check on Self-Distillation in Agentic Reinforcement Learning The paper benchmarks ten self-distillation methods in agentic reinforcement learning across varied model sizes, tasks, and computational costs. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Text-based agent environments, not multimodal models. | |
| JetSpec++: Foresight Parallel Drafting with Self Refinements The paper introduces a speculative decoding system combining self-refinement and pipelined drafting to reduce wall-clock generation time. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Applies to autoregressive language models, not multimodal. | |
| Evolve Together, Fall into Place: How Social Scaffolding Shapes Self-Evolving Multi-Agent Systems The paper investigates how team composition and organizational interdependence affect performance in self-evolving multi-agent systems. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Does not involve multimodal models. | |
| Delta Tokenizer: Reference-Identical Tokenization of Session Deltas and Segment-Parallelism for Long Prompts on Multicore CPUs The paper introduces a CPU front-end tokenizer that encodes only session deltas to speed up multi-turn inference. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Text tokenization, not multimodal. | |
| Model vs. User: Whose Values Prevail? Inference-Time Value Alignment in Multi-Turn Conversations The paper studies the durability of inference-time value steering in multi-turn conversations when user and model values conflict. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Text-based conversational alignment, not multimodal. | |
| MAPLE: A Multi-Axis Parametric Knowledge Extraction Lens for LLM Evaluation The paper presents a diagnostic evaluation framework to measure how inference-time techniques affect parametric knowledge extraction in models. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Text-only evaluation, not multimodal. | |
| EGCForest: Structured Agent Memory Organization for Enterprise Group Chats The paper proposes a multi-view hierarchical memory framework to improve information retrieval from enterprise group chats. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Text-based chat retrieval, not multimodal. | |
| OpenTWBench (Open Taiwan Benchmark): Taiwan-Domain Knowledge Evaluation for Large Language Models, and How the Measurement Apparatus Shapes the Score The paper introduces a two-part benchmark to evaluate language models on general and examination knowledge concerning Taiwan. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on text-based knowledge benchmarks, not multimodal models. | |
| RCCM: Diagnosing Gains and Harms in Reasoning-Prompt Transfer The paper analyzes question-level variations and prompt-execution mismatches when transferring reasoning prompts across different language models. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Text-based reasoning, not multimodal. | |
| Beyond Benchmark Accuracy: A Diagnostic Test for LLM Chemical Prediction The paper proposes a diagnostic test hiding target identity to evaluate structure-property chemical reasoning in language models. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on chemical structure reasoning, not multimodal VLMs. | |
| Not All Criteria Matter: Importance-Aware Criteria Weighting for Rubric-Based LLM Evaluation The paper introduces a method that weights rubric criteria through pairwise model comparisons to improve evaluation accuracy. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on text evaluation, not multimodal models. | |
| MolHop: Scaffold-Hopping Molecular Design with Agentic Reinforcement Learning The paper presents an agentic reinforcement learning framework that navigates organized chemical spaces for cross-scaffold molecular discovery. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on molecular graphs and RL, not multimodal foundation models. | |
| MemDuet: Learning Dual-Process Memory for Long-Context Reasoning The paper introduces a dual-process memory framework coordinating small and large models for efficient long-context reasoning. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Does not involve multimodal data. | |
| To Reason but Not to Stop: Understanding and Recovering Pruned Large Reasoning Models The paper shows pruned reasoning models fail by looping, proposing an activation steering method to restore proper termination. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Text-only reasoning models, not multimodal. | |
| Choosing the Number of Attention Heads from Intrinsic Dimension: A Small-Scale Study The paper tests a heuristic to set the number of transformer attention heads using the intrinsic dimension of token representations. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Does not involve multimodal models. | |
| SoK: Grading Open-Text Legal Answers, from Law School Exams to Trustworthy Rubric Judges The paper systematizes grading requirements for open-text legal exams and compares eight evaluation paradigms using human and model judges. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Applies to text legal examinations, not multimodal inputs. | |
| Gray-Box Hallucination Detection in LLMs with Token Identities The paper introduces a lightweight gray-box hallucination detector combining token probabilities with vector representations of token identities. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Operates on text token probabilities without multimodal data. | |
| LintEval: Benchmarking LLMs on Linter-Guided File Refactoring The paper presents a multilingual benchmark to evaluate how language models resolve linter alerts while preserving program functionality. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Deals with source code in C++, Python, and Java. | |
| What Is a Clarification? Estimating Model-Dependent Input Ambiguity The paper formalizes clarification criteria to estimate model-dependent input ambiguity based on accessible and valid answers. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on text QA and SQL ambiguities, not multimodality. | |
| Language Exposure Matters: Distribution and Order in Multilingual BabyLM Pre-training The paper evaluates how language exposure distribution and ordering affect the performance of multilingual models trained with limited data. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on text pre-training, not multimodal or vision tasks. | |
| Sparse Autoencoders for Mechanistic Interpretability in NLP: A Critical Survey of Recoverability, Interpretability, and Utility This paper surveys 138 studies on sparse autoencoders in NLP, examining dictionary design, feature recoverability, and practical utility. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on language model feature representations, not multimodal models. | |
| Gap Detection: Mitigating Homogenization through Demand and Defensibility This paper introduces Gap Detection, an optimization method that identifies unoccupied semantic directions to reduce homogenization in open-ended model answers. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Deals with text embedding spaces and language generation. | |
| RM-OPD: On-Policy Distillation for Generative Reward Models This paper presents RM-OPD, a framework that trains domain-specialized reward models via on-policy distillation into a generalist model. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Targets text preference reward models, not multimodal tasks. | |
| TreeSpark: Learning to Draft the Trees You Verify for Speculative Decoding This paper introduces TreeSpark, a framework training semi-autoregressive drafters to generate candidate trees aligned with speculative verification. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on text LLM speculative decoding, not multimodal models. | |
| When Occupation-derived Skills Don't Help: On the Limits of Skill-Augmented Career Path Prediction This paper analyzes career path prediction systems, reporting that explicit skill representations provide marginal improvement over text-only job descriptions. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Uses textual job descriptions, without vision or speech modalities. | |
| Scaling Multilingual Mixture-of-Experts: How Language Coverage Changes Capacity and Compute Allocation This paper proposes a joint scaling law evaluating how adding experts in multilingual mixture-of-experts models affects language capacity and loss penalties. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on multilingual text scaling, not cross-modal processing. | |
| Neural POS Tagging for the Low-Resource Saraiki This paper introduces an annotated part-of-speech tagging dataset and evaluates pretrained transformer encoders for the low-resource Saraiki language. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Addresses text-only sequence tagging in Saraiki script. | |
| Taking the PULSE of Medical Reasoning: Step-Level Retrieval Timing for Knowledge-Grounded Inference This paper proposes PUlSE, using internal layer representations to detect factual errors and trigger step-level retrieval during medical reasoning. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Operates on text-based medical QA without vision or audio. | |
| MMLA-RSI: Recursive Self-Improvement through RTT and RTMA This paper proposes MMLA-RSI, a recursive self-improvement framework using reasoning-time policy training and authoritative memory adaptation. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Does not address vision or cross-modal modalities. | |
| Do AI Agents Gossip? Evaluative Talk in Human and AI-Agent Online Communities This paper investigates whether gossip discourse naturally emerges in autonomous AI agent online communities compared to human discussion boards. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Investigates text-based forum agents, not multimodal models. | |
| Lexical Persistence Without Semantic Entrenchment in AI Agent Communities This paper measures semantic drift and stability across AI agent communities, finding higher drift and lower temporal continuity than human forums. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Studies forum text and lexical persistence, with no multimodal modalities. | |
| From Triage to Impact: A Multi-Hop Reasoning Benchmark for Explainable Persian Hate Speech This paper introduces a multi-hop reasoning benchmark to evaluate fine-grained, explainable hate speech analysis in Persian. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Analyzes text-based social media hate speech, with no multimodal input. | |
| FOLD: Position-Free Far Memory with Local Order for Long-Context Speculative Decoding This paper introduces FOLD, a training-free speculative decoding approach treating distant tokens as position-free far memory for long-context generation. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Evaluates long-context text decoding, not multimodal image/audio systems. | |
| The COUNT: Counting N-grams Unlocks Scaling on Host Memory and Training-Free Adaptation This paper introduces COUNT, scaling language model memory without training parameters by storing precomputed n-gram counts directly in host memory. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses purely on text pretraining and n-grams, not multimodal data. | |
| Same Claim, Different Interpretations: How LLM Agents Resolve Ambiguity in Scientific Claims This paper proposes Claim Cards to make agent interpretations explicit when verifying ambiguous natural-language scientific claims. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Evaluates textual scientific claims and evidence, without visual or multimodal components. | |
| MolLean: A Process-Verifiable Language for Chemical Reasoning This paper introduces MolLean, a verifiable language and tool interface that grounds chemical reasoning in formal state transitions and explicit checks. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on chemical domain-specific languages and text reasoning, not multimodal models. | |
| ChronoCulture: Evaluating Cultural Awareness Across Time in Large Language Models This paper introduces ChronoCulture to evaluate how well language models capture temporal changes in cultural attitudes across 72 countries. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Benchmarking focuses on textual survey questions, containing no multimodal modalities. | |
| Lost in Translation: Why Lower Bias Scores Don't Mean Fairer Models This paper evaluates multilingual bias audits across ten language models, showing that lower stereotype scores often reflect translation artifacts rather than fairer behavior. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Investigates text-only multilingual translation bias without multimodal data. | |
| ReaDy: Recurrent States as Dynamic Context for Speculative Drafting in Hybrid Linear Attention Large Language Models This paper presents ReaDy, a speculative decoding method using hybrid linear attention recurrent states to provide position-specific context during drafting. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Applies to autoregressive language model decoding without multimodal inputs or outputs. | |
| Manacá-1B-Instruct-2: Reproducible Safety Alignment for a Native Brazilian-Portuguese LLM. Negative DPO Results, a Controlled Composition Scan, and a Silent Failure Mode of the LLM Judge Documents safety alignment methods, negative preference tuning outcomes, and automated judge limitations for a Portuguese language model. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses strictly on Portuguese text alignment, with no multimodal components. | |
| Latent Compaction with Test-Time Training Compresses long-context key-value caches using test-time training on generic document prompts without task-specific labels. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on text long-context KV cache compression, not multimodal models. | |
| The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery Presents an end-to-end distributed system orchestrating coding agents to execute computational research experiments alongside human scientists. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on coding and scientific orchestration, not multimodal vision-language models. | |
| Evaluating Large Language Models for Multilingual Abstractive Summarization: A Comprehensive Study of English and Bengali Evaluates closed and open models on English and Bengali abstractive summarization using lexical metrics, factuality checks, and automated judges. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Evaluates text-only summarization without any visual, audio, or multimodal components. | |
| Who Should We Listen to More? Welfare-Aware Preference Acquisition for Pluralistic Alignment Introduces a welfare-aware objective to allocate limited human feedback across diverse individuals when training pluralistic alignment models. Preprint | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Focuses on text alignment preferences without multimodal data. | |
| Graph Coarsening Enables Escaping Local Optima in Compute-Constrained Nonconvex MAX-CUT Combines spectral graph coarsening with continuous relaxation to find high-quality solutions for compute-constrained MAX-CUT problems. Peer-reviewed | Multimodal models via arxiv · 15 Oct, 15:06 | Not relevant Does not involve multimodal language or vision models. |
901–960 of 1,924← PreviousNext →