Every paper checked — nothing hidden

All papers

Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.

All17,313 In brief24 More (on-topic)3,333 Not relevant13,754 Not assessed87 Older352 Could not confirm93 Blogs with no paper266 All blogs284

Paper	Topic	Relevance
How Much Audio Is Left In An Embedding? An Inversion Audit Of Audio Encoders Audits audio encoders by reconstructing music using a latent diffusion model to quantify signal properties preserved across training objectives. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Audits audio representation retention via latent diffusion, not LLM evaluation.
How Much Audio Is Left In An Embedding? An Inversion Audit Of Audio Encoders Audits audio encoders by reconstructing music using a latent diffusion model to quantify signal properties preserved across training objectives. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Investigates representation fidelity in audio encoders, not deployment safety or guardrails.
How Much Audio Is Left In An Embedding? An Inversion Audit Of Audio Encoders Audits audio encoders by reconstructing music using a latent diffusion model to quantify signal properties preserved across training objectives. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Investigates audio encoder feature retention, not LLM agents.
Syn-Omni: Structured Specialization and Progressive Collaboration for Omnimodal Embeddings Proposes Syn-Omni, a framework combining modality-expert LoRA paths and progressive routing to balance shared and modality-specific omnimodal embeddings. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Presents a model architecture evaluated across tasks, not an LLM benchmark study.
Syn-Omni: Structured Specialization and Progressive Collaboration for Omnimodal Embeddings Proposes Syn-Omni, a framework combining modality-expert LoRA paths and progressive routing to balance shared and modality-specific omnimodal embeddings. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not examine AI safety, alignment, or deployment reliability.
Syn-Omni: Structured Specialization and Progressive Collaboration for Omnimodal Embeddings Proposes Syn-Omni, a framework combining modality-expert LoRA paths and progressive routing to balance shared and modality-specific omnimodal embeddings. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Studies multimodal embeddings rather than retrieval-augmented generation pipelines.
Syn-Omni: Structured Specialization and Progressive Collaboration for Omnimodal Embeddings Proposes Syn-Omni, a framework combining modality-expert LoRA paths and progressive routing to balance shared and modality-specific omnimodal embeddings. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses on omnimodal representation learning, not autonomous LLM agents.
Syn-Omni: Structured Specialization and Progressive Collaboration for Omnimodal Embeddings Proposes Syn-Omni, a framework combining modality-expert LoRA paths and progressive routing to balance shared and modality-specific omnimodal embeddings. Preprint	Computer vision for industrial applications via arxiv · 12 Oct, 16:28	Not relevant Focuses on general multimodal embeddings across vision/audio, not industrial vision tasks.
DVD: Dynamic Vector Decoding for Efficient MLLM-based Perception Introduces Dynamic Vector Decoding to efficiently map 2D and 3D perception coordinates into compact discrete tokens for multimodal LLMs. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Proposes a perception tokenization scheme, not multi-step language model agents.
DVD: Dynamic Vector Decoding for Efficient MLLM-based Perception Introduces Dynamic Vector Decoding to efficiently map 2D and 3D perception coordinates into compact discrete tokens for multimodal LLMs. Preprint	Computer vision for industrial applications via arxiv · 12 Oct, 16:28	Not relevant Evaluates general 2D/3D perception benchmarks, not industrial inspection or manufacturing.
DVD: Dynamic Vector Decoding for Efficient MLLM-based Perception Introduces Dynamic Vector Decoding to efficiently map 2D and 3D perception coordinates into compact discrete tokens for multimodal LLMs. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Focuses on perceptual decoding architecture, not language model evaluation.
DVD: Dynamic Vector Decoding for Efficient MLLM-based Perception Introduces Dynamic Vector Decoding to efficiently map 2D and 3D perception coordinates into compact discrete tokens for multimodal LLMs. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not address safety, guardrails, or model alignment.
Unlocking the Regulatory Genome by ARGUS: An Evidence-Constrained Agentic Framework for Interpreting Single Nucleotide Variants Proposes ARGUS, an agentic framework separating deterministic genomic models from LLM reasoning to interpret noncoding single nucleotide variants reliably. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Does not address model compression, distillation, or efficient inference.
Unlocking the Regulatory Genome by ARGUS: An Evidence-Constrained Agentic Framework for Interpreting Single Nucleotide Variants Proposes ARGUS, an agentic framework separating deterministic genomic models from LLM reasoning to interpret noncoding single nucleotide variants reliably. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Applies agents to genomic analysis, not AI deployment safety or alignment.
Unlocking the Regulatory Genome by ARGUS: An Evidence-Constrained Agentic Framework for Interpreting Single Nucleotide Variants Proposes ARGUS, an agentic framework separating deterministic genomic models from LLM reasoning to interpret noncoding single nucleotide variants reliably. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Focuses on genomic interpretation framework rather than general LLM evaluation benchmarks.
Machine Learning Meets High-Energy Nuclear Physics: From Pattern Recognition to Physics-Integrated Discovery Surveys recent developments in physics-integrated machine learning applications across high-energy nuclear physics experiments and simulations. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not focus on LLM safety, reliability guardrails, or alignment.
Machine Learning Meets High-Energy Nuclear Physics: From Pattern Recognition to Physics-Integrated Discovery Surveys recent developments in physics-integrated machine learning applications across high-energy nuclear physics experiments and simulations. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Physics domain review, not an LLM benchmark or evaluation study.
Machine Learning Meets High-Energy Nuclear Physics: From Pattern Recognition to Physics-Integrated Discovery Surveys recent developments in physics-integrated machine learning applications across high-energy nuclear physics experiments and simulations. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Surveys machine learning in nuclear physics, not autonomous agent workflows.
Multi-Agent Egocentric World Model with Fine-Grained Embodied Interaction Presents ME-World, an egocentric world model generating synchronized multi-agent first-person video streams with fine-grained physical interactions. Preprint	Computer vision for industrial applications via arxiv · 12 Oct, 16:28	Not relevant Focuses on multi-agent egocentric video generation, not industrial manufacturing or inspection.
Multi-Agent Egocentric World Model with Fine-Grained Embodied Interaction Presents ME-World, an egocentric world model generating synchronized multi-agent first-person video streams with fine-grained physical interactions. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not evaluate safety, red teaming, or operational guardrails.
Multi-Agent Egocentric World Model with Fine-Grained Embodied Interaction Presents ME-World, an egocentric world model generating synchronized multi-agent first-person video streams with fine-grained physical interactions. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Introduces consistency metrics specifically for its own generative world model, not general LLM benchmarking.
Learning Probabilistic Logic Programs with Functional Gradient Guided Language Models Introduces grasp, a framework combining functional gradient boosting with LLM proposal generation to learn probabilistic logic programs from relational data. Peer-reviewed	AI agents via arxiv · 12 Oct, 16:28	Not relevant Uses LLM as a proposal generator in functional gradient boosting, not autonomous agentic systems.
Learning Probabilistic Logic Programs with Functional Gradient Guided Language Models Introduces grasp, a framework combining functional gradient boosting with LLM proposal generation to learn probabilistic logic programs from relational data. Peer-reviewed	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not focus on safety, red teaming, jailbreaks, or deployment reliability.
Learning Probabilistic Logic Programs with Functional Gradient Guided Language Models Introduces grasp, a framework combining functional gradient boosting with LLM proposal generation to learn probabilistic logic programs from relational data. Peer-reviewed	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Introduces a neurosymbolic method for logic programs rather than an LLM evaluation framework.
Looking Inside LLMs: Small-World Connectivity as a Signature of Reasoning Performance Analyzes attention-head activation graphs to show that small-world connectivity correlates with reasoning performance in large language models. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Analyzes internal attention graph metrics, not tool use or autonomous agents.
Looking Inside LLMs: Small-World Connectivity as a Signature of Reasoning Performance Analyzes attention-head activation graphs to show that small-world connectivity correlates with reasoning performance in large language models. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on internal attention network topology rather than safety, alignment, or guardrails.
A Structural Theory of Cognitive Representation and Problem Solving,Contexts, Invariance, and the Knowledge Space Proposes a formal structural framework defining context formation, abstraction, and invariance operations for cognitive representation and problem solving. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Theoretical cognitive representation paper; does not address deployment safety or guardrails.
A Structural Theory of Cognitive Representation and Problem Solving,Contexts, Invariance, and the Knowledge Space Proposes a formal structural framework defining context formation, abstraction, and invariance operations for cognitive representation and problem solving. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Proposes theoretical cognitive representations without evaluating or benchmarking LLMs.
A Structural Theory of Cognitive Representation and Problem Solving,Contexts, Invariance, and the Knowledge Space Proposes a formal structural framework defining context formation, abstraction, and invariance operations for cognitive representation and problem solving. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Presents a theoretical cognitive framework, not LLM-based autonomous agent architectures.
Verdict Without the Rule: Diagnosing and Auditing Regulatory Rule Sensitivity in LLM Compliance Systems Evaluates whether LLM compliance systems adjust verdicts when governing rules are perturbed, finding many verdicts are insensitive to rule changes. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Studies rule sensitivity in compliance classification, not multi-step tool-using language model agents.
Verdict Without the Rule: Diagnosing and Auditing Regulatory Rule Sensitivity in LLM Compliance Systems Evaluates whether LLM compliance systems adjust verdicts when governing rules are perturbed, finding many verdicts are insensitive to rule changes. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Focuses on regulatory compliance decision sensitivity, not retrieval systems or RAG pipelines.
Verdict Without the Rule: Diagnosing and Auditing Regulatory Rule Sensitivity in LLM Compliance Systems Evaluates whether LLM compliance systems adjust verdicts when governing rules are perturbed, finding many verdicts are insensitive to rule changes. Preprint	LLMs for code via arxiv · 12 Oct, 16:28	Not relevant Does not evaluate code generation, repair, or software engineering tasks.
Verdict Without the Rule: Diagnosing and Auditing Regulatory Rule Sensitivity in LLM Compliance Systems Evaluates whether LLM compliance systems adjust verdicts when governing rules are perturbed, finding many verdicts are insensitive to rule changes. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Does not address model compression, distillation, quantization, or efficient serving.
Verdict Without the Rule: Diagnosing and Auditing Regulatory Rule Sensitivity in LLM Compliance Systems Evaluates whether LLM compliance systems adjust verdicts when governing rules are perturbed, finding many verdicts are insensitive to rule changes. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Focuses entirely on text-based regulatory compliance without multimodal inputs.
Prior or Feedback? What an LLM Uses When Adapting Neural Operators Investigates whether LLM scientific agents rely on task priors or experimental feedback when configuring neural operator adaptations. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Presents an experimental analysis of agent decision drivers rather than an LLM benchmark.
Prior or Feedback? What an LLM Uses When Adapting Neural Operators Investigates whether LLM scientific agents rely on task priors or experimental feedback when configuring neural operator adaptations. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Does not focus on model compression, distillation, or efficient serving.
Prior or Feedback? What an LLM Uses When Adapting Neural Operators Investigates whether LLM scientific agents rely on task priors or experimental feedback when configuring neural operator adaptations. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Examines adaptation feedback mechanisms rather than deployment safety, jailbreaking, or guardrails.
Prediction-Powered Data Fusion for Treatment Effect Estimation Develops a framework fusing small randomized controlled trials with large observational data to improve treatment effect estimation precision without bias. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not address AI safety guardrails, hallucination, or alignment in deployment.
Prediction-Powered Data Fusion for Treatment Effect Estimation Develops a framework fusing small randomized controlled trials with large observational data to improve treatment effect estimation precision without bias. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Addresses statistical treatment effect estimation, not LLM agents.
Prediction-Powered Data Fusion for Treatment Effect Estimation Develops a framework fusing small randomized controlled trials with large observational data to improve treatment effect estimation precision without bias. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Does not cover model compression, efficient inference, or small language models.
Prediction-Powered Data Fusion for Treatment Effect Estimation Develops a framework fusing small randomized controlled trials with large observational data to improve treatment effect estimation precision without bias. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Focuses on causal inference estimators rather than LLM benchmarks or evaluation methods.
RiCo: Neural Simulation of Rigid-Body Interactions via Local Contact Reasoning Presents RiCo, a neural physics simulation method that models rigid-body interactions via sparse local contact surface reasoning. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Proposes a physics simulation method, not an LLM benchmark.
RiCo: Neural Simulation of Rigid-Body Interactions via Local Contact Reasoning Presents RiCo, a neural physics simulation method that models rigid-body interactions via sparse local contact surface reasoning. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Focuses on rigid-body physics modeling rather than vision-language foundation models.
RiCo: Neural Simulation of Rigid-Body Interactions via Local Contact Reasoning Presents RiCo, a neural physics simulation method that models rigid-body interactions via sparse local contact surface reasoning. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not explore AI safety, reliability guardrails, or alignment.
RiCo: Neural Simulation of Rigid-Body Interactions via Local Contact Reasoning Presents RiCo, a neural physics simulation method that models rigid-body interactions via sparse local contact surface reasoning. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses on physical contact simulation rather than language model agents.
RiCo: Neural Simulation of Rigid-Body Interactions via Local Contact Reasoning Presents RiCo, a neural physics simulation method that models rigid-body interactions via sparse local contact surface reasoning. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Addresses contact simulation efficiency, not small language models or LLM quantization.
RiCo: Neural Simulation of Rigid-Body Interactions via Local Contact Reasoning Presents RiCo, a neural physics simulation method that models rigid-body interactions via sparse local contact surface reasoning. Preprint	Computer vision for industrial applications via arxiv · 12 Oct, 16:28	Not relevant Deals with general physics simulation rather than industrial visual inspection or defect detection.
ContiLNN: Mitigating Slice Sampling Discontinuity with Liquid Neural Networks for Medical Image Restoration Proposes ContiLNN, which uses closed-form continuous-time modules to improve cross-slice continuity and restoration in 2D medical imaging. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Presents an image restoration network architecture, not an evaluation benchmark.
ContiLNN: Mitigating Slice Sampling Discontinuity with Liquid Neural Networks for Medical Image Restoration Proposes ContiLNN, which uses closed-form continuous-time modules to improve cross-slice continuity and restoration in 2D medical imaging. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Deals with medical imaging reconstruction, not LLM agents or tool use.
ContiLNN: Mitigating Slice Sampling Discontinuity with Liquid Neural Networks for Medical Image Restoration Proposes ContiLNN, which uses closed-form continuous-time modules to improve cross-slice continuity and restoration in 2D medical imaging. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Addresses volumetric medical image restoration rather than multimodal text-image foundation models.
ContiLNN: Mitigating Slice Sampling Discontinuity with Liquid Neural Networks for Medical Image Restoration Proposes ContiLNN, which uses closed-form continuous-time modules to improve cross-slice continuity and restoration in 2D medical imaging. Preprint	Computer vision for industrial applications via arxiv · 12 Oct, 16:28	Not relevant Applies to medical clinical imaging rather than industrial manufacturing inspection or robotics vision.
ContiLNN: Mitigating Slice Sampling Discontinuity with Liquid Neural Networks for Medical Image Restoration Proposes ContiLNN, which uses closed-form continuous-time modules to improve cross-slice continuity and restoration in 2D medical imaging. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not investigate AI safety, guardrails, or alignment.
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. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not study safety alignment, jailbreaking, or deployment guardrails.
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. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Does not explore document retrieval, grounding, or RAG systems.
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. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Evaluates textual/code game environments, not multimodal vision-language processing.
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. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Does not focus on model compression, quantization, or lightweight serving.
Overcoming Prior Barriers: Supervised Fine-Tuning under Long-Tail Distribution Analyzes long-tail concept learning during supervised fine-tuning using a theoretical prior barrier metric to guide data needs. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Does not investigate model pruning, quantization, or small on-device models.
Overcoming Prior Barriers: Supervised Fine-Tuning under Long-Tail Distribution Analyzes long-tail concept learning during supervised fine-tuning using a theoretical prior barrier metric to guide data needs. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Focuses on SFT data selection for LLMs, not retrieval-augmented generation architectures.
Overcoming Prior Barriers: Supervised Fine-Tuning under Long-Tail Distribution Analyzes long-tail concept learning during supervised fine-tuning using a theoretical prior barrier metric to guide data needs. Preprint	LLMs for code via arxiv · 12 Oct, 16:28	Not relevant Addresses general concept fine-tuning, not specialized code generation or software repair.
Overcoming Prior Barriers: Supervised Fine-Tuning under Long-Tail Distribution Analyzes long-tail concept learning during supervised fine-tuning using a theoretical prior barrier metric to guide data needs. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Does not address deployment safety, jailbreak robustness, or guardrails.

16,681–16,740 of 17,313← Previous Next →