Every paper checked — nothing hidden
All papers
Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.
All17,312In brief24More (on-topic)3,333Not relevant13,753Not assessed87Older351Could not confirm93Blogs with no paper266All blogs284
| Paper | Topic | Relevance | Feedback |
|---|---|---|---|
| Omni-Modal LLMs Know More Than They Say This paper analyzes why omni-modal language models perceive non-textual inputs correctly yet fail to mention them in text outputs. Ranked #99New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Investigates representation alignment across vision, audio, IMU, and text in omni-modal LLMs. | |
| TextVerifier: Seeing Beyond Recognition to Verify Chinese Text Rendering The paper introduces TEXTVERIFIER, a verification framework using character structural sequences to detect fine rendering defects in generated Chinese text. Ranked #100New testPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Evaluates multimodal models and verifiers on identifying detailed visual text rendering defects. | |
| Diagnosing and Correcting Reward Hacking in Verifiable Instruction Following The paper analyzes reward hacking in verifiable instruction following and introduces a likelihood bonus method to maintain answer quality. Ranked #100Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 15:06 | On-topic Mitigates reward hacking and maintains response quality in verifiable instruction following. | |
| Ontology-Based Post-Coordinated Expression Assembly: from Clinical Narrative to Queryable Knowledge Graphs This paper introduces UNIGRAPHE, a deterministic method to assemble clinical text into queryable knowledge graphs conforming to SNOMED CT grammar. Ranked #100Preprint0/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | On-topic Converts unstructured clinical text into queryable knowledge graphs for structured evidence retrieval. | |
| Machine--human agreement in marking undergraduate assignments is specific to model, prompt and institution The paper examines machine grading on undergraduate essays across three universities, finding models compress score distributions and vary across institutions. Ranked #100Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Evaluates the reliability and cross-context validity of LLM judges marking assignments. | |
| Humans vs. Machines: A Comparison With AI Agents in Software Reverse Engineering This paper benchmarks language model agents against skilled humans on binary software reverse engineering tasks using interactive tools. Ranked #100Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates autonomous LLM agents interacting with reverse engineering tools over multi-step tasks. | |
| SpecForge: Recursive Self-Improvement for Coding Agents via Behavioral Specifications The paper presents a self-improvement method where coding agents refine natural-language behavioral specifications using execution-verified tests. Ranked #100New way to see a problemPreprint6/10 suggested | LLMs for code via arxiv · 15 Oct, 06:17 | On-topic Enhances coding agents through recursive self-improvement using behavioral code specifications. | |
| DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoning The paper introduces a training-free contrastive decoding method where a small model guides a larger model toward shorter reasoning chains. Ranked #100Surprising resultPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Proposes training-free contrastive decoding to reduce token usage and improve inference efficiency. | |
| PrismaCoder: Pushing the Boundaries of Open Multimodal Code Intelligence through Decomposed Rubric Synthesis Introduces a decomposed rubric synthesis method to create training data for open multimodal models that generate visual code artifacts. Ranked #101New testPreprint6/10 suggested | LLMs for code via arxiv · 15 Oct, 15:06 | On-topic Develops multimodal code generation for visual web artifacts and code. | |
| Rethinking LLM-Guided Visual Question Answering: Does Deeper Knowledge Integration Really Help? The study compares LLM guidance integration strategies for visual question answering, showing shallow late fusion outperforms deeper attention injection. Ranked #101Surprising resultPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Investigates integrating LLM knowledge into vision-language architectures for visual question answering. | |
| Scaling Repeated-Data Pretraining: Batch Size, Regularization, and GPU-Time Tradeoffs This paper investigates batch size, regularization methods, and runtime tradeoffs when pretraining language models on repeated finite data. Ranked #101Surprising resultPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Explores trade-offs between batch size, regularization, and GPU compute efficiency during pretraining. | |
| CultureBot: A Framework for Retrieval-Augmented Exploration of Cultural Heritage Data This paper introduces an interactive retrieval-augmented framework that combines knowledge graphs and text retrieval to explore digital cultural heritage archives. Ranked #101Preprint0/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | On-topic Introduces an interactive framework combining knowledge graphs and retrieval-augmented generation for cultural collections. | |
| Beyond Belief Nesting: Rethinking Higher-Order Theory of Mind in Large Language Models The paper introduces HopToM, a benchmark testing whether higher-order theory-of-mind tasks in language models genuinely evaluate nested reasoning rather than shortcuts. Ranked #101Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Uncovers flaws in higher-order Theory of Mind tests and introduces HopToM, a certified benchmark for perspective reasoning. | |
| MCPDisCo: Seesaw-Mitigated Tuning for MCP-Native Agents This paper presents a reinforcement learning framework to train language model agents across tools and servers in the Model Context Protocol ecosystem. Ranked #101New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Proposes a reinforcement learning framework for agents using Model Context Protocol tools and servers. | |
| ProbabilityBench: Model Evaluation and Calibration with Exact Ground-Truth Uncertainty The paper introduces ProbabilityBench to evaluate model uncertainty calibration against analytically computed, ground-truth instance-level probabilities across three text environments. Ranked #101New testPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Focuses on measuring true model uncertainty and calibration reliability in decision-making. | |
| Memory as Weights: Internalizing Long-Term History for Streaming Videos The paper compresses video history into evolving lightweight model weights instead of competing for limited input context length. Ranked #102New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 17 Oct, 06:17 | On-topic Develops a vision-language framework and dataset specifically for long-term streaming video understanding. | |
| DivAlgo: A systematic evaluation of algorithmic diversity for code generation models The paper introduces a benchmark and similarity metric to measure the algorithmic diversity of solutions produced by code generation models. Ranked #102Surprising resultPreprint6/10 suggested | LLMs for code via arxiv · 15 Oct, 15:06 | On-topic Introduces a benchmark evaluating LLM competence and algorithmic diversity in code generation. | |
| KhyontekAI at CheckThat! 2026: Multilingual Dense Retrieval and Cross-Encoder Reranking for Scientific Web Claims The authors present a multilingual dense retrieval and reranking framework to link scientific web claims to their source publications. Ranked #102Peer-reviewed0/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | On-topic Proposes dense retrieval and cross-encoder reranking methods for grounding claims. | |
| Connecting the Dots: Enduring Performance Gaps in ASR Systems Between Standard and Minority English Varieties The paper tracks automatic speech recognition performance from 2021 to 2025, finding persistent accuracy disparities between standard American English and minority accents. Ranked #102Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Conducts a longitudinal benchmark evaluation revealing persistent accent disparities and benchmark limitations in ASR systems. | |
| Rubrics from the Task, Verdicts from Evidence: Pointwise Rubric Rewards with Tool-Augmented Judges This paper proposes a rubric-based reward model that derives criteria strictly from instructions and evaluates responses pointwise using tool assistance. Ranked #102Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Improves reward models and evaluation rubrics used for aligning LLM outputs. | |
| Unconstrained CTI-REALM with Open-Weights This paper evaluates open-weight language models against closed frontier models on a cybersecurity detection benchmark using an interactive telemetry environment. Ranked #102Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates language model agents querying telemetry stores to solve threat intelligence tasks. | |
| ShellElastic: Scalable Elastic Pretraining with Simultaneous Maximal Updates and Hyperparameter Transfer Introduces an elastic transformer pretraining architecture enabling simultaneous hyperparameter transfer and subnetwork execution within a single model. Ranked #102New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Introduces elastic supernet architectures enabling variable-budget sub-networks for efficient inference. | |
| CoTranslator: Reverse to Forward Chain-of-Thought Translator for Reasoning Distillation The paper proposes a method to translate reverse reasoning paths into verified forward chains of thought for distillation. Ranked #103New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 17 Oct, 06:17 | On-topic Presents a method to distill reasoning capabilities into smaller student models. | |
| LEASH: Uncertainty-Aware Weak-to-Strong Escalation and Oversight for Computer-Use Agents The paper proposes a training-free framework that uses uncertainty signals to escalate computer-use tasks from weak models to strong models. Ranked #103New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 17 Oct, 06:17 | On-topic Provides runtime uncertainty monitoring and oversight to prevent agent failures in deployment. | |
| HERO: Harnessing Evidence for Test-time Refinement and Prompt Optimization in Program Generation The paper proposes a method combining offline execution-verified prompt updates and test-time refinement to improve program generation in language models. Ranked #103Surprising resultPreprint6/10 suggested | LLMs for code via arxiv · 15 Oct, 15:06 | On-topic Optimizes prompts and refines code generation for SQL, Cypher, and Python. | |
| Simple Linear Corrections for Compression in Late Interaction The paper proposes a linear score correction method to recover retrieval quality in compressed late-interaction document representations. Ranked #103Preprint0/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | On-topic Addresses scoring corrections and compression for late-interaction dense retrieval models. | |
| Label Collapse in Diffusion Language Model In-Context Learning This paper analyzes label collapse failure modes across diffusion and autoregressive language models during in-context learning classification. Ranked #103Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Exposes an undetected failure mode in in-context learning evaluation metrics. | |
| When Does an ASR Error Matter? Task-Role Diagnostics for Tool-Using Voice Agents The paper introduces a diagnostic protocol to evaluate how speech recognition errors affect structured tool calling in voice agents. Ranked #103New testPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Analyzes how speech recognition errors impact tool-calling behavior in voice agents. | |
| Jag: Direct Box Prediction for Efficient Visual Grounding Adapts multimodal models for direct continuous bounding box prediction using a lightweight regression head in a single forward pass. Ranked #103New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 12 Oct, 16:28 | On-topic Adapts pretrained multimodal vision-language models for direct visual grounding and coordinate localization. | |
| HPCAgent-Bench: A Framework for Sound Benchmarking of Code-Optimizing Agents, Harnesses, and Skills The authors introduce a multi-platform benchmark to measure the performance and tool efficacy of code-optimizing agents on high-performance computing kernels. Ranked #104New testPreprint6/10 suggested | LLMs for code via arxiv · 17 Oct, 06:19 | On-topic Presents a sound benchmark framework for code-optimizing agents across hardware targets. | |
| WorldMemBench: A Comprehensive Benchmark for Memory of Video World Models WorldMemBench introduces an evaluation benchmark assessing static entity retention and dynamic event persistence in video world models over multi-revisit paths. Ranked #104New testPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Evaluates memory and temporal consistency in generative video world models. | |
| Fin_Agent at FinEval 2026: Task-Aware Methods for Multilingual Financial NLP The paper describes a system submission combining rules and language models for multilingual financial named-entity recognition and question answering. Ranked #104Preprint0/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | On-topic Implements retrieval-augmented generation using BGE-M3 and Qwen3-32B for financial QA. | |
| Rule vs.Consequence: Dissociable Internal Representations of Moral Reasoning in LLMs This paper uses activation steering and probing to identify distinct internal representations for rule-based versus consequence-based moral reasoning in language models. Ranked #104New way to see a problemPeer-reviewed7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | On-topic Evaluates LLM moral reasoning using a 2x2 factorial framing and probing framework. | |
| Scaling Inference-Time Compute with Teams of Coding Agents The paper studies compute allocation across multi-agent coding teams to identify Pareto-optimal trade-offs between cost and performance. Ranked #104Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Explores coordinating multi-agent teams and allocating inference-time compute across planning and execution. | |
| Beyond Perfect Teammates: Multi-Agent Collaboration with Unreliable Colleagues This paper introduces UnCoBench to benchmark how language model agents handle unhelpful, incomplete, or missing communications from partner agents. Ranked #104New testPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Evaluates system robustness and failure recovery in multi-agent deployments. | |
| Agentic-TTT: Training test-time policy for test-time training Proposes a policy mechanism enabling language models to decide when, how, and which test-time training procedures to execute. Ranked #104New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Learns a policy to dynamically trade off test-time compute against utility improvements. | |
| Benchmarking and Enhancing Design-Aware Code Review in the Linux Kernel The paper presents a knowledge-augmented framework and benchmark to help automated code review account for historical software design decisions. Ranked #105New testPreprint6/10 suggested | LLMs for code via arxiv · 17 Oct, 06:19 | On-topic Introduces a benchmark and method for evaluating coding agents on kernel code review. | |
| Video-Arena: A Scalable Agentic Video Benchmark for Spatiotemporal Capability This paper introduces a video benchmark derived from competitive game records to evaluate dynamic spatiotemporal world state tracking. Ranked #105New testPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Presents a benchmark evaluating spatiotemporal world state tracking in video models. | |
| Strong yet Controllable LLM Abstention without Learning to Abstain This paper proposes TAPS, combining standard accuracy training with post-hoc confidence selection to give controllable abstention rates without specialized tuning. Ranked #105Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | On-topic Analyzes metrics decomposing risk into answering and distinguishing abilities across QA benchmarks. | |
| The Curse of Knowledge in Role-Playing Language Models The paper tests whether role-playing language models leak private or unobserved character information across simulated interactive scenes. Ranked #105Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates information boundaries and hidden state tracking in multi-agent and role-playing simulations. | |
| Looking Complete Is Not Being Grounded: Evaluating LLM Sitreps Against Human-Authored Situation Reports This paper evaluates language-model-generated disaster situation reports against human-written reports using a domain-specific sixteen-slot information schema. Ranked #105Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Measures factual grounding and hallucination risks in LLM-generated reports for high-stakes settings. | |
| Specialists Learn Faster: Knowledge Organization in Pre-training Shapes Knowledge Update Efficiency Examines how chronological versus shuffled pre-training knowledge sequencing influences the sample efficiency of subsequent model knowledge updates. Ranked #105Surprising resultPreprint6/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:29 | On-topic Studies efficient knowledge update strategies during training to minimize update steps and compute cost. | |
| Retrieval-Augmented Multimodal Language Model with Heterogeneous Evidence Integration for Clinical Prognosis Proposes PRISM, a multimodal clinical prognostic model combining patient records with retrieved population-level embedding memory banks. Ranked #105Preprint0/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28 | On-topic Incorporates population-level representation retrieval from training cohorts to ground clinical predictive outputs. | |
| RelLedger: Relational Information Governance for LLM Agent Organizations This paper introduces RelLedger, an auditing framework to trace how relational histories drive reporting choices and actions in multi-agent systems. Ranked #106New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | On-topic Introduces an audit framework to evaluate information sharing behavior in agent organizations. | |
| DuoDiscover: Scaling Autonomous Discovery Beyond the Evaluation Bottleneck The paper proposes an agent framework that discovers lightweight proxy evaluators to bypass costly benchmark evaluations during autonomous engineering discovery. Ranked #106New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Presents a multi-agent system of autonomous coding agents for real-world engineering discovery. | |
| Probable but Not Valid: A Psychometric Audit of Decision Models as Synthetic Survey Respondents This paper audits whether decision models and generative language models can serve as valid synthetic respondents for psychometric surveys. Ranked #106Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Audits the reliability and validity of using language models for synthetic human profiling. | |
| EventLens: Event-Structure Reinforcement Learning for Video Understanding This paper presents EventLens, a reinforcement learning framework that trains multimodal models to understand video through explicit event structure recovery. Ranked #106New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Focuses on temporal event structure understanding in multimodal vision-language models. | |
| Compile the Table: Query-Calibrated Operator Compression for Tabular In-Context Learning Compresses tabular in-context learning key-value caches into query-calibrated prototypes to improve throughput without sacrificing accuracy. Ranked #106New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:29 | On-topic Focuses on KV cache compression and efficient inference throughput for tabular in-context learning. | |
| Verifiable Retrieval for Legal Reasoning through Structure-driven Knowledge Graph The paper evaluates knowledge-graph-informed chunking and retrieval strategies over German legal code to support verifiable retrieval-augmented generation. Ranked #106Preprint0/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28 | On-topic Improves RAG chunking, navigation, and citation accuracy using structure-driven knowledge graphs. | |
| Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition Introduces AAArena, a benchmark evaluating how AI agents revise executable code policies across 12 adversarial games under fixed budgets. Ranked #106New testPreprint5/10 suggested | LLMs for code via arxiv · 12 Oct, 16:28 | On-topic Evaluates agents synthesizing, analyzing, and revising executable code policies across matches. | |
| Equal CER, Unequal Damage: Severity-Weighted Character Error Rate for OCR Proposes a severity-weighted character error rate that charges text transcription errors based on their practical impact on reader understanding. Ranked #107New testPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Evaluates OCR transcription performance in vision-language models. | |
| Who Sets the Pace of Judge Hacking? Judge Capability versus the Policy-Judge Gap This work analyzes whether policy capability gaps or judge models dictate reward hacking when training models against language-model judges. Ranked #107Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | On-topic Investigates judge hacking dynamics and the limits of LLM-as-a-judge. | |
| BRIDGE: Selective Boundary Repair with Dependency Graph Evidence in Multi-Agent Systems The paper introduces BRIDGE, a runtime intervention framework that detects and repairs broken information handoffs across dependency graphs in multi-agent workflows. Ranked #107New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Introduces a runtime control framework to detect and repair broken handoffs between agents in multi-agent systems. | |
| Judges Follow the Reference: The Derivation Gap in Verifying Tool-Call Arguments The authors identify a derivation gap where LLM judges reject correct tool arguments lacking direct textual references, and propose a dedicated reference model. Ranked #107Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Addresses reliability and verification of tool-call arguments in agent deployments. | |
| Spectral Weight Decay: Inducing Low-Rank Structure in Neural Network Weights Introduces a spectral weight decay update inducing low-rank weight matrices to improve SVD-based post-training model compression. Ranked #107Surprising resultPreprint6/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:29 | On-topic Induces low-rank weight structure directly enabling post-training SVD compression and faster GPU inference for LLMs. | |
| Natural Language to First-Order Logic LLM-based Autoformalization The paper reviews LLM-based first-order logic autoformalization, distinguishing ontology extraction from logical translation and outlining key benchmarking challenges. Ranked #107New way to see a problemPreprint5/10 suggested | LLMs for code via arxiv · 12 Oct, 16:28 | On-topic Surveys autoformalization translating natural language into formal logic programs and specifications. | |
| CliQ-RRG: Clinical-Knowledge Guided Disease-aware Visual-Textual Alignment for QA-Style Radiology Report Generation Presents a two-stage visual-textual alignment and retrieval-augmented framework for generating structured radiology reports from medical scans. Ranked #107Preprint0/10 suggested | Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28 | On-topic Integrates GraphRAG retrieval to ground radiology report generation in clinical knowledge. | |
| MultiAlloc: Budget-Aware Precision Allocation for KV Cache Quantization The paper presents a profiling-guided framework to allocate precision budgets adaptively across layers and tokens during KV cache quantization. Ranked #108Surprising resultPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Proposes budget-aware precision allocation for KV cache quantization to improve LLM inference efficiency. | |
| MemForm: Benchmarking Memory Formation Poisoning in LLM Agents The paper introduces MemForm to benchmark adversarial memory formation poisoning across five input interfaces in language model agents. Ranked #108New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | On-topic Introduces a diagnostic benchmark for memory formation poisoning. | |
| Reliable Memory from Unstable Growth: Curating Self-Written Agent Playbooks with Randomized Experiments Presents a randomized experiment framework to test and curate self-written agent playbooks, preventing task failure caused by unverified lessons. Ranked #108New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Develops a randomized experimentation method to curate self-written playbooks for multi-step LLM agents. |
721–780 of 3,333← PreviousNext →