Every paper checked — nothing hidden
All papers
Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.
All17,313In brief24More (on-topic)3,333Not relevant13,754Not assessed87Older352Could not confirm93Blogs with no paper266All blogs284
| Paper | Topic | Relevance | Feedback |
|---|---|---|---|
| Learning Speech Attribute Representations via Multi-Teacher Distillation Distills multiple speech attributes from four teacher models into a single shared speech encoder. Ranked #338Preprint0/10 suggested | Small and efficient models via arxiv · 17 Oct, 06:17 | On-topic Consolidates multiple speech representations into a single shared encoder via distillation. | |
| Agreement versus Softmax under Spatial Disambiguation This paper evaluates calibration differences between object-level softmax and agreement metrics when resolving ambiguous 3D referring queries. Ranked #338Surprising resultPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Investigates confidence calibration under spatial disambiguation in 3D referring vision-language models. | |
| On the Monitorability of Natural Language Autoencoders: Identification vs Discrimination This paper uses referential communication games to analyze why natural language autoencoders fail to provide discriminative oversight of model activations. Ranked #338Surprising resultPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Investigates the reliability and limits of natural language autoencoders for model oversight. | |
| Words to World: Benchmarking LLMs in Interactive Physical Environments Introduces a closed-loop benchmark testing language models controlling physical robotic hardware and reacting to real sensor feedback. Ranked #338New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Benchmarks language models operating closed-loop physical systems, invoking tools, and planning actions. | |
| IsoVRBench: Video Reward Modeling Without Human Feedback The paper presents a benchmark of synthetic paired comparisons to evaluate video reward models across specific visual and motion dimensions. Ranked #338New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces IsoVRBench, a benchmark for evaluating video reward models. | |
| Batch-Paced: Fixed-Budget Batch Decoding in Discrete Diffusion Language Models The paper proposes Batch-Paced, a method that dynamically allocates decoding compute across batched requests to improve discrete diffusion language model inference. Ranked #339Preprint0/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Proposes fixed-budget batch decoding to optimize compute allocation and improve inference efficiency in diffusion LLMs. | |
| From Hindsight to Foresight: Dissecting Clinical Evidence Acquisition in Large Language Models This paper analyzes dynamic clinical evidence gathering in language models, identifying failures in evidence identification, updating, and diagnostic reasoning. Ranked #339New way to see a problemPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Investigates systematic reliability failures and dynamic decision breakdown in clinical LLM deployments. | |
| ACED-Bench: Evaluating How LLM Agents Acquire and Act on Causal Evidence The paper introduces a neuro-symbolic benchmark to evaluate whether language model agents actively investigate environments and make causal decisions. Ranked #339New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Focuses directly on evaluating how autonomous LLM agents actively acquire evidence, intervene, and make decisions. | |
| EarthAlign: Learning Language-Aligned Representations of Earth System Data The paper releases a multimodal dataset and dual-encoder model to align natural language descriptions with global Earth system meteorological data. Ranked #339New testPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Learns joint multimodal representations aligning meteorological numerical data with natural language. | |
| VL-DocIR: A Benchmark for Vision-Based Long Document Information Retrieval Introduces a visual document retrieval benchmark focused on multi-page and cross-document evidence configurations in long documents. Ranked #339New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces VL-DocIR, a new benchmark for evaluating vision-based long-document retrieval. | |
| PixARG: Pixel-Asynchronous Retrieval-Augmented Diffusion for Fast and Realistic Image Generation Integrates real-world image retrieval into diffusion models using pixel-level affinity maps and heterogeneous noise initialization to accelerate generation. Ranked #340New way to see a problemPeer-reviewed4/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Combines text prompts and retrieved images for multimodal conditioned diffusion generation. | |
| Scaling Multi-Agent Intelligence with Substrate-Centric Coordination ScaleMAS coordinates multi-agent systems using an asynchronous shared datastore with limited search to improve scaling and fault tolerance. Ranked #340New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:18 | On-topic Designs a shared datastore architecture to scale multi-agent collaboration and prevent failure. | |
| The Image Still Has a Say: Final-Answer Repair for Multimodal Chain-of-Thought The paper repairs multimodal chain-of-thought final answers using original image features without altering the generated reasoning trace. Ranked #340New way to see a problemPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Reduces visual reasoning hallucinations by anchoring final answers back to image evidence. | |
| Prefilling-dLLM v2: Learned Chunk Routing for Long-Context Diffusion Language Models The paper proposes a training approach that uses learned chunk routing to speed up long-context diffusion language model processing. Ranked #340Preprint0/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Proposes learned chunk routing to enable efficient long-context inference in diffusion language models. | |
| StableJudge: A Lightweight Adversarial Defense Framework for LLM-as-a-Judge The paper presents StableJudge, an external lightweight sentry model that sanitizes prompt-injection attacks before LLM-as-a-judge scoring. Ranked #340New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Protects LLM-as-a-judge scoring frameworks from score-hijacking prompt injection attacks. | |
| SPROUT: Scaling Agentic Reinforcement Learning for Long-Horizon Software Engineering The paper presents SPROUT, a snapshot-based rollout framework to scale reinforcement learning for long-horizon software engineering tasks. Ranked #341Surprising resultPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:18 | On-topic Develops a snapshot-based rollout framework for long-horizon coding agents. | |
| Decouple to Forget: Semantic Mediation for Sensitive Unlearning in Large Language Models The paper proposes an unlearning method that removes sensitive facts from model responses while preserving surrounding non-sensitive factual context. Ranked #341New way to see a problemPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Presents a method to remove sensitive factual data to prevent information leakage in deployment. | |
| More Reader Data or a Better Teacher? Updating a Model of Crowd Highlights in Late 2026 The paper compares direct prompting, model-distilled supervision, and larger reader datasets for predicting user text highlighting behavior on web pages. Ranked #341Preprint0/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Evaluates whether smaller 3B models trained on user data can outperform larger zero-shot or teacher LLMs. | |
| CREATE: Creativity Evaluation of VLMs using the Visual Remote Associates Test This paper introduces a visual associative test benchmark to evaluate the lateral reasoning and creative association capabilities of vision-language models. Ranked #341New testPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Evaluates vision-language models on visual Remote Associates puzzles connecting multiple images. | |
| LLM-based User-Product Relevance Evaluation for Dynamic Product Advertising The paper introduces a large language model-based metric for offline evaluation of user-product relevance in dynamic product advertising retrieval. Ranked #341New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Proposes an LLM-as-a-judge metric and framework for evaluating retrieval systems. | |
| CoLA: Multi-VLA Coordination via Learned Adapters This paper introduces learned communication adapters and action heads to coordinate multiple vision-language-action robotic policies during collaborative tasks. Ranked #342Surprising resultPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:18 | On-topic Studies multi-agent coordination and communication between vision-language-action policies in shared workspaces. | |
| Beyond Sparse Captions: Aligning Slide-Level Text and Patch-Level Vision in Pathology LIMIT aligns slide-level pathology reports with patch-level vision features by freezing the vision encoder and training only the text encoder. Ranked #342New way to see a problemPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Aligns clinical text reports with microscopic image features in vision-language pretraining. | |
| Perturbation, Not Provenance: Why Activation Steering Does Not Make a Robust Text Watermark The paper tests activation steering as a text watermarking technique, finding that standard logit-based watermarking provides more robust detection against paraphrasing. Ranked #342Surprising resultPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Evaluates the robustness of text watermarking techniques against paraphrasing attacks for safe deployment provenance. | |
| Teacher-Guided Tool Learning for Small Language Models via Bidirectional KL Constraints Proposes a teacher-guided reinforcement learning method using bidirectional constraints to help small language models learn effective tool use. Ranked #342Preprint0/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Enables small language models to acquire complex tool-use capabilities using teacher guidance and RL. | |
| VisExMEM: Visual Evidence-Based Explainable Multimodal Evaluation Metric Presents a reference-free evaluation metric that decomposes captions into atomic claims and verifies them against visual evidence. Ranked #342New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces a claim-level, explainable evaluation metric and benchmark for image-text assessment. | |
| PAWA: Panoramic Action-conditioned World Modeling for Multiple Agents The paper introduces a dataset and framework pairing synchronized point-of-view and panoramic video to evaluate multiplayer action-conditioned world models. Ranked #343New way to see a problemPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Introduces a multimodal panoramic video and action dataset and modeling framework. | |
| Mind Your Own Role: Reducing Role Overstepping for Effective and Efficient Multi-Agent Collaboration via Dynamic Role Activation Steering Applies inference-time activation steering to keep multi-agent systems from overstepping their assigned roles. Ranked #343Surprising resultPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:18 | On-topic Prevents role overstepping in multi-agent LLM systems using dynamic activation steering at inference time. | |
| A Bethe-Ansatz Source of Early Token Correlations: A marginal-preserving construction for parallel proposals This paper constructs joint token probability laws from physics models to improve speculative verification in parallel language model decoding. Ranked #343Preprint0/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Presents theoretical foundations for parallel token proposals in speculative decoding. | |
| SLS-Bench: A Benchmark for Incident Log Summarization with Synthetic Observability Data Introduces a benchmark of synthetic observability data and incident reports to evaluate automated incident log summarization. Ranked #343New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Presents SLS-Bench, a new benchmark dataset for evaluating incident summarization. | |
| Automated Assembly Instruction Generation from CAD Models Using Grounded Large Language Models: A Human-in-the-Loop Framework Builds a human-in-the-loop framework using grounded language models to automatically generate natural-language assembly instructions from CAD models. Ranked #343New way to see a problemPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:29 | On-topic Focuses on human-in-the-loop validation and grounding to prevent errors in manufacturing deployment. | |
| Does Failure Feedback Direct Agent Self-Modification? Evaluates whether failure feedback helps self-improving agents fix specific target bugs rather than general errors. Ranked #344Surprising resultPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 06:18 | On-topic Investigates whether failure feedback helps self-modifying LLM agents repair specific target errors effectively. | |
| A Production-Grade, High-Efficiency Agentic RL Framework for Frontier LLMs with Validated Recipes This paper introduces LoongSage, an asynchronous reinforcement learning framework designed to run multi-turn agent tasks with reduced system overhead. Ranked #344Preprint0/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Focuses on high-throughput, efficient distributed training and inference infrastructure for models. | |
| Empathy Is Not the Scapegoat for Declines in Factuality and Reliability Examines whether model empathy degrades reliability, showing that warm behavior effects depend on tasks and training approaches rather than an inevitable trade-off. Ranked #344Surprising resultPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Investigates whether empathy training compromises model factuality and reliability in stressful deployment scenarios. | |
| HyWe: Unifying Search and Operation for Real-World Hybrid Web Tasks Introduces a 35B model, trajectory pipeline, and benchmark unifying search and browser interaction within a single web agent. Ranked #344New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces HyWeBench to evaluate web agent performance on complex interactive tasks. | |
| CineMem: Attention-Aligned Fine-grained Memory Retrieval for Multi-Shot Long Video Generation Introduces CineMem, an attention-aligned memory retrieval framework using clustered KV caches for autoregressive multi-shot long video generation. Ranked #344New way to see a problemPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Addresses video generation and multimodal long-context modeling using attention-aligned memory retrieval. | |
| AVC-Reward: Event-Centric Audio-Video Consistency Evaluation The authors present AVC-Reward, an event-centric framework to evaluate temporal, narrative, and attribute consistency in generated audio-video content. Ranked #345New testPreprint4/10 suggested | Multimodal models via arxiv · 17 Oct, 06:17 | On-topic Evaluates cross-modal consistency between audio and video generated streams. | |
| Co-Evolving Acting and Forecasting for Language Agents Introduces multi-step action forecasting targets alongside actor rollouts to train language agents more effectively from intermediate trajectory steps. Ranked #345New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 15:06 | On-topic Trains language agents to plan and act over multi-step horizons using forecasting targets. | |
| DeepThink: A Conceptual Model for Knowledge-Addressable Recurrent Language Modeling This paper proposes DeepThink, a conceptual architecture that decouples knowledge capacity, activation width, and recurrent depth in language models. Ranked #345Preprint0/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Proposes conditional computation and modular recurrent architecture to balance capacity and execution depth. | |
| How Reliable Are System-One Decisions in Medicine? A Comparative Study of Jev, Jev-like Models, and LLMs Evaluates fast, system-one decision models and language models across fifteen medical datasets, testing their accuracy, format robustness, and abstention awareness. Ranked #345Surprising resultPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Examines decision robustness, reversal rates, and failure to abstain in high-stakes clinical deployments. | |
| Learning to GUIDE the Group: Self-Learning LLM Facilitation Across Multi-Party Tasks Presents a framework for developing language model facilitators that balance group dynamics, psychological safety, and task goals across multi-party conversations. Ranked #345New way to see a problemPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces an evaluation framework assessing facilitator models on objective and subjective metrics. | |
| ReferFlow: Beyond Final-Target Localization for Complex Referring Expressions This paper introduces ReferFlow-Bench and an accompanying method to evaluate and perform multi-step trajectory-level referring expression comprehension. Ranked #346New testPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Studies trajectory-level visual grounding across multi-step multimodal referring expressions. | |
| Maximising Transparency in Context Compaction Improves Retention of Facts and Commitments for Lightweight Coding Agents The paper presents a deterministic context compaction method designed to preserve factual history and commitments in lightweight coding agents. Ranked #346New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 15:06 | On-topic Investigates context compaction techniques for lightweight autonomous coding agents. | |
| ASSET: Adaptive Speech Segmentation via Reinforcement Learning The paper introduces ASSET, a reinforcement learning method that adaptively compresses speech encoder outputs into continuous audio tokens. Ranked #346Preprint0/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Compresses acoustic encoder token length for efficient inference in LLMs. | |
| ToolHACE: A Dataset for Detecting Hallucinations in Tool-Based LLM Responses This paper introduces a benchmark dataset to evaluate hallucination detection in language model responses after external tool execution. Ranked #346New testPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Addresses the detection of hallucinations and reliability failures in deployed tool-using models. | |
| Query Your Way Out: Benchmarking Multi-Turn Knowledge Reasoning through Generalized Puzzle Solving This paper introduces the Active Probing Benchmark to evaluate multi-turn active questioning and specialized domain reasoning in language models. Ranked #346Surprising resultPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces a multi-turn benchmark to evaluate knowledge reasoning and active probing. | |
| Recognizing Dialogue Ambiguity via Candidate Set Visual Grounding This paper introduces a task and benchmark for tracking candidate sets of referent objects across ambiguous interactive dialogue turns. Ranked #347New testPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Studies vision-language models grounding ambiguous dialogue into visual candidate sets. | |
| SWE-Nexus: Automated Composition of Software Engineering Tasks from Test Execution The paper introduces an automated pipeline that builds repository-level software engineering benchmarks from test execution footprints. Ranked #347New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 15:06 | On-topic Uses multi-agent construction pipelines and evaluates software engineering agents. | |
| 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 #347Preprint0/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Focuses on compressing document vectors to make retrieval cheaper and more efficient. | |
| Towards Universal Calibration for In-Context Classification This paper presents a universal calibrator to improve confidence estimates of in-context classification without requiring model-specific fine-tuning. Ranked #347New way to see a problemPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Improves calibration to prevent overconfident erroneous predictions in deployment. | |
| RetryBench: Separating Retry Permission from Recovery Outcomes in GPT Tool Use The paper presents a benchmark evaluating how language model agents handle unauthorized tool call retries after receiving timeouts. Ranked #347New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces RetryBench to evaluate agent failure handling and tool retry behaviors. | |
| ELARSpeech: A 30K-Hour Dataset Across 675 Underrepresented Languages for Long-Tail Multilingual Speech Technologies The paper introduces a speech dataset spanning 675 underrepresented languages to support speech recognition and text-to-speech technologies. Ranked #348New testPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Introduces a massive audio-speech dataset and models spanning speech and text across 675 languages. | |
| Memorizing Who, Knowing How: Benchmarking Personal and Procedural Memory for Persistent Assistants The paper introduces a benchmark to evaluate how personal assistants handle interdependent personal and procedural memories over long trajectories. Ranked #348New testPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 15:06 | On-topic Benchmarks persistent personal agents managing long-term user memory. | |
| Lexical Redefinition Reveals Context-Sensitive Instruction Following This paper tests how language models handle explicit instructions when subsequent contextual sentences provide conflicting word definitions. Ranked #348Surprising resultPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Analyzes how easily instructions can be overridden or weakened by subsequent contextual evidence. | |
| Google Earth AI Geospatial Reasoning Agent: Capabilities and a Systematic Evaluation The paper presents Google Earth AI's geospatial reasoning agent along with evaluation benchmarks for complex multi-step geospatial workflows. Ranked #348New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Releases GeoExpertQueries and GeoKTools benchmarks to systematically evaluate geospatial agent workflows. | |
| Adaptive DHiSparse: Speculative Decoding over a Load-Adaptive KV Hierarchy Combines load-adaptive hierarchical key-value caching with speculative decoding to accelerate long-context language model serving. Ranked #348Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Introduces load-adaptive KV memory hierarchy and speculative decoding to optimize long-context LLM serving efficiency. | |
| Anuvani-Diarization: Zero Shot Speaker Diarization and Annotation Conventions Across 22 Languages This paper presents Anuvani-Diarization, evaluating zero-shot speaker diarization across 22 Indian languages with uncertainty-aware silver supervision. Ranked #349New testPreprint4/10 suggested | Multimodal models via arxiv · 15 Oct, 15:06 | On-topic Focuses on multi-speaker audio processing across diverse Indian languages. | |
| One Monitor Does Not Fit All: Formalism-Agnostic Runtime Verification for Agentic AI This paper develops a runtime verification architecture that decouples monitoring formalisms from execution boundaries to safely control AI agents. Ranked #349New way to see a problemPreprint6/10 suggested | AI agents via arxiv · 15 Oct, 15:06 | On-topic Proposes a runtime verification architecture for monitoring agent tool use and delegation. | |
| Estimating Uncertainty of Omnimodal Large Language Models via External Omnimodal Embeddings This paper develops an uncertainty estimation framework for omnimodal models using external embeddings to score reasoning grounding and coherence. Ranked #349New way to see a problemPreprint6/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Improves deployment reliability by estimating confidence and grounding of multimodal responses. | |
| CoverSynth: Coverage-Guided Generation of Retrieval Evaluation Sets This paper introduces a synthetic data generation and assessment framework designed to produce comprehensive evaluation benchmarks for enterprise retrieval systems. Ranked #349New testPreprint6/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces an assessment framework and generator for enterprise retrieval evaluation sets. | |
| Not All Dimensions Are Equal: Outlier-Aware Attention Linearization with Grouped Quadratic Kernels Presents Outlier-Aware Attention Linearization, using quadratic approximations to retain outlier dimensions during efficient Transformer attention evaluation. Ranked #349Preprint0/10 suggested | Small and efficient models via arxiv · 12 Oct, 16:28 | On-topic Linearizes softmax attention using quadratic kernels to achieve efficient, linear-time Transformer inference. |
1,981–2,040 of 3,333← PreviousNext →