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
Resource-Efficient 3D Shape and Orientation Estimation of Road Users using Roadside RGB Camera and 3D Radar The authors propose a resource-efficient two-stage approach for estimating 3D road user shapes and orientations from roadside cameras and radar. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Computer vision perception model, not an LLM benchmark.
Resource-Efficient 3D Shape and Orientation Estimation of Road Users using Roadside RGB Camera and 3D Radar The authors propose a resource-efficient two-stage approach for estimating 3D road user shapes and orientations from roadside cameras and radar. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Focuses on intelligent transportation systems, not industrial inspection or defect detection.
Agaram 2.5B: The First Tamil-First Language Model Trained from Scratch with Tool Calling. A Playbook for Under-Represented Languages The authors introduce a 2.5-billion-parameter bilingual Tamil-English language model trained from scratch with tool-calling capabilities. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Not focused on retrieval-augmented generation or external document retrieval.
Agaram 2.5B: The First Tamil-First Language Model Trained from Scratch with Tool Calling. A Playbook for Under-Represented Languages The authors introduce a 2.5-billion-parameter bilingual Tamil-English language model trained from scratch with tool-calling capabilities. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Not focused on code generation or software engineering tasks.
Agaram 2.5B: The First Tamil-First Language Model Trained from Scratch with Tool Calling. A Playbook for Under-Represented Languages The authors introduce a 2.5-billion-parameter bilingual Tamil-English language model trained from scratch with tool-calling capabilities. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Text-only language model, not multimodal.
Agaram 2.5B: The First Tamil-First Language Model Trained from Scratch with Tool Calling. A Playbook for Under-Represented Languages The authors introduce a 2.5-billion-parameter bilingual Tamil-English language model trained from scratch with tool-calling capabilities. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Training playbook for low-resource languages, not an evaluation methodology.
Agaram 2.5B: The First Tamil-First Language Model Trained from Scratch with Tool Calling. A Playbook for Under-Represented Languages The authors introduce a 2.5-billion-parameter bilingual Tamil-English language model trained from scratch with tool-calling capabilities. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Mentions basic tool calling in passing; main focus is language model training.
Agaram 2.5B: The First Tamil-First Language Model Trained from Scratch with Tool Calling. A Playbook for Under-Represented Languages The authors introduce a 2.5-billion-parameter bilingual Tamil-English language model trained from scratch with tool-calling capabilities. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Not focused on AI safety, red-teaming, or guardrails.
FedGraph: Defending Federated Large Language Model Fine-Tuning Against Backdoor Attacks via Graph-Based Aggregation The authors present a graph-based aggregation method to defend federated language model fine-tuning against backdoor attacks by identifying update consistency patterns. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Presents a defense framework, not an LLM benchmark or evaluation study.
FedGraph: Defending Federated Large Language Model Fine-Tuning Against Backdoor Attacks via Graph-Based Aggregation The authors present a graph-based aggregation method to defend federated language model fine-tuning against backdoor attacks by identifying update consistency patterns. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on federated learning defenses, not agent planning or execution.
How Well Do LLMs Perform in Peer Review? Evidence from Finance and Accounting The authors evaluate language models on finance peer review tasks, finding predictive ability for publication outcomes through correctness rather than originality assessments. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies LLMs evaluating papers, not multi-step agent planning or tool use.
STAMP as a topology-aware molecular language for fragment-level learning and editing The authors introduce a topology-aware molecular representation language that enables fragment-level representation learning and precise structural edits. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Proposes molecular representation for chemistry, not LLM agent systems.
STAMP as a topology-aware molecular language for fragment-level learning and editing The authors introduce a topology-aware molecular representation language that enables fragment-level representation learning and precise structural edits. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Computational chemistry paper, not AI safety or deployment reliability.
STAMP as a topology-aware molecular language for fragment-level learning and editing The authors introduce a topology-aware molecular representation language that enables fragment-level representation learning and precise structural edits. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Focuses on molecular editing representations, not LLM evaluation.
The Singularity Has a Distribution: Exact Timing Laws and Detection Limits for a Stochastic Intelligence Explosion The authors formulate a stochastic model to analyze the timing distributions and detection limits of potential recursive artificial intelligence explosions. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Theoretical singularity modeling rather than practical deployment safety or reliability.
The Singularity Has a Distribution: Exact Timing Laws and Detection Limits for a Stochastic Intelligence Explosion The authors formulate a stochastic model to analyze the timing distributions and detection limits of potential recursive artificial intelligence explosions. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Theoretical timing analysis rather than LLM benchmarking or evaluation.
The Singularity Has a Distribution: Exact Timing Laws and Detection Limits for a Stochastic Intelligence Explosion The authors formulate a stochastic model to analyze the timing distributions and detection limits of potential recursive artificial intelligence explosions. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Theoretical mathematical modeling of recursive self-improvement, not AI agents.
Thinker-WM: A Unified World Model with Shared Multimodal Representations for Embodied Intelligence The authors propose an embodied world model that unifies RGB video, metric depth, and robot actions into a shared dynamic representation. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Proposes a world model architecture, not a benchmark or evaluation method.
Thinker-WM: A Unified World Model with Shared Multimodal Representations for Embodied Intelligence The authors propose an embodied world model that unifies RGB video, metric depth, and robot actions into a shared dynamic representation. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Focuses on embodied robotics, not industrial manufacturing defect detection.
Thinker-WM: A Unified World Model with Shared Multimodal Representations for Embodied Intelligence The authors propose an embodied world model that unifies RGB video, metric depth, and robot actions into a shared dynamic representation. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Not focused on safety guardrails, jailbreaks, or deployment risks.
Autonomous Science Bench: Evaluating AI Agents for Autonomous Scientific Discovery The authors introduce a benchmark to evaluate autonomous scientific discovery agents making iterative experimental decisions under fixed budgets. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Benchmarks capability and discovery rather than safety guardrails or alignment.
Learning Perturbation-Specific Representations through Conditional Energy Prior for Cellular Response Prediction The authors present an energy-based prior for cellular perturbation models to better distinguish specific therapeutic signals from general biological background noise. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Not focused on model compression, quantization, or efficient inference.
Learning Perturbation-Specific Representations through Conditional Energy Prior for Cellular Response Prediction The authors present an energy-based prior for cellular perturbation models to better distinguish specific therapeutic signals from general biological background noise. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Domain-specific biological predictive model, not an LLM benchmark.
Learning Perturbation-Specific Representations through Conditional Energy Prior for Cellular Response Prediction The authors present an energy-based prior for cellular perturbation models to better distinguish specific therapeutic signals from general biological background noise. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Applies energy priors to cellular response prediction, not LLM agents.
Learning Perturbation-Specific Representations through Conditional Energy Prior for Cellular Response Prediction The authors present an energy-based prior for cellular perturbation models to better distinguish specific therapeutic signals from general biological background noise. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Drug discovery representation learning, not AI deployment safety.
Readout Scaling Shapes Sparse Feature Recovery The authors analyze readout scaling in sparse autoencoders and propose canonicalized coordinates to improve the recovery of underlying generative factors. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Not focused on coding assistants or software engineering.
Readout Scaling Shapes Sparse Feature Recovery The authors analyze readout scaling in sparse autoencoders and propose canonicalized coordinates to improve the recovery of underlying generative factors. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Mechanistic interpretability study, not LLM benchmark evaluation.
Readout Scaling Shapes Sparse Feature Recovery The authors analyze readout scaling in sparse autoencoders and propose canonicalized coordinates to improve the recovery of underlying generative factors. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Does not examine multimodal or vision-language models.
Readout Scaling Shapes Sparse Feature Recovery The authors analyze readout scaling in sparse autoencoders and propose canonicalized coordinates to improve the recovery of underlying generative factors. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies sparse autoencoder interpretability, not agent workflows.
Readout Scaling Shapes Sparse Feature Recovery The authors analyze readout scaling in sparse autoencoders and propose canonicalized coordinates to improve the recovery of underlying generative factors. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Does not involve retrieval-augmented generation.
Readout Scaling Shapes Sparse Feature Recovery The authors analyze readout scaling in sparse autoencoders and propose canonicalized coordinates to improve the recovery of underlying generative factors. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Investigates feature recovery in SAEs, not model compression or efficiency.
MemoryVLN: From Ego-Aware Memory Representation to Cross-Context Memory Reuse in Vision-Language Navigation The authors propose a memory-driven vision-language navigation system using temporal compression and geometric-semantic representations to support long-horizon trajectory tasks. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Not focused on AI safety, alignment, or guardrails.
MemoryVLN: From Ego-Aware Memory Representation to Cross-Context Memory Reuse in Vision-Language Navigation The authors propose a memory-driven vision-language navigation system using temporal compression and geometric-semantic representations to support long-horizon trajectory tasks. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Focuses on embodied navigation, not industrial inspection or defect detection.
MemoryVLN: From Ego-Aware Memory Representation to Cross-Context Memory Reuse in Vision-Language Navigation The authors propose a memory-driven vision-language navigation system using temporal compression and geometric-semantic representations to support long-horizon trajectory tasks. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Introduces a navigation architecture, not an evaluation framework or benchmark.
ReTerm: Recovery Training for Self-Improving Terminal Agents The authors present a training framework and dataset that enables terminal coding agents to self-improve by recovering from execution errors. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Centers on agent task performance rather than safety guardrails or alignment.
ReTerm: Recovery Training for Self-Improving Terminal Agents The authors present a training framework and dataset that enables terminal coding agents to self-improve by recovering from execution errors. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Introduces a training method, not a benchmark or evaluation study.
ReTerm: Recovery Training for Self-Improving Terminal Agents The authors present a training framework and dataset that enables terminal coding agents to self-improve by recovering from execution errors. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Fine-tunes a model rather than advancing model efficiency or compression.
Task-evoked eye movements as candidate digital biomarkers of depression severity in adolescents and young adults The authors analyze task-evoked eye movements in adolescents to identify objective behavioral markers corresponding to depression severity scores. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Clinical study of eye-movement biomarkers, not LLM agents.
Task-evoked eye movements as candidate digital biomarkers of depression severity in adolescents and young adults The authors analyze task-evoked eye movements in adolescents to identify objective behavioral markers corresponding to depression severity scores. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Medical depression diagnosis study, not AI deployment safety or reliability.
Task-evoked eye movements as candidate digital biomarkers of depression severity in adolescents and young adults The authors analyze task-evoked eye movements in adolescents to identify objective behavioral markers corresponding to depression severity scores. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Not an evaluation or benchmark of language models.
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. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Does not involve multimodal reasoning or vision-language models.
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. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Evaluates tool arguments, not retrieval-augmented generation 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. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Does not address model compression, quantization, or efficient serving.
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. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Not focused on code generation or software engineering agents.
Aligning Faithfulness for Theory-Scale Autoformalization The authors propose evaluating autoformalization faithfulness by aligning phrase-level correspondences between informal mathematics and formal code under contextual interpretations. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Does not involve multimodal models or vision-language data.
Aligning Faithfulness for Theory-Scale Autoformalization The authors propose evaluating autoformalization faithfulness by aligning phrase-level correspondences between informal mathematics and formal code under contextual interpretations. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Does not study retrieval-augmented generation systems.
Aligning Faithfulness for Theory-Scale Autoformalization The authors propose evaluating autoformalization faithfulness by aligning phrase-level correspondences between informal mathematics and formal code under contextual interpretations. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on mathematical autoformalization, not LLM agents or tool use.
Aligning Faithfulness for Theory-Scale Autoformalization The authors propose evaluating autoformalization faithfulness by aligning phrase-level correspondences between informal mathematics and formal code under contextual interpretations. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on mathematical autoformalization accuracy, not deployment safety.
Aligning Faithfulness for Theory-Scale Autoformalization The authors propose evaluating autoformalization faithfulness by aligning phrase-level correspondences between informal mathematics and formal code under contextual interpretations. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Does not address model compression, pruning, or efficient inference.
EgoOmniWorld: Unifying Egocentric Understanding via Ego-Action-World Interactions The authors introduce a benchmark evaluating egocentric video understanding structured around an interactive loop of user perception, action, and world feedback. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Evaluates video interaction QA, not safety risks or guardrails.
EgoOmniWorld: Unifying Egocentric Understanding via Ego-Action-World Interactions The authors introduce a benchmark evaluating egocentric video understanding structured around an interactive loop of user perception, action, and world feedback. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on egocentric video QA, not LLM agentic tool use.
EgoOmniWorld: Unifying Egocentric Understanding via Ego-Action-World Interactions The authors introduce a benchmark evaluating egocentric video understanding structured around an interactive loop of user perception, action, and world feedback. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Deals with egocentric human video, not industrial defect detection.
The Smart Buildings Challenge: Learning Foundation Models for Real-World Optimization at Scale The authors introduce a challenge benchmark to develop foundation models for building HVAC energy optimization that transfer across different physical structures. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on HVAC energy efficiency, not LLM deployment safety.
The Smart Buildings Challenge: Learning Foundation Models for Real-World Optimization at Scale The authors introduce a challenge benchmark to develop foundation models for building HVAC energy optimization that transfer across different physical structures. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on HVAC optimization policies, not LLM agent systems.
The Smart Buildings Challenge: Learning Foundation Models for Real-World Optimization at Scale The authors introduce a challenge benchmark to develop foundation models for building HVAC energy optimization that transfer across different physical structures. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Presents a challenge for building HVAC control, not LLM evaluation.
A Global View of How LLMs Value Human Time The paper evaluates how commercial language models value human time by measuring recommended delay compensation amounts across 193 different countries. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Evaluates geographical valuation bias in LLMs, not multi-step agent architectures.
The Geometry of Language Model Logits: Generic Structure and Learned Departures The paper analyzes the geometric distribution of language model output logits, identifying a consistent log-linear relationship that emerges during pretraining. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Examines logit distribution geometry, not agent planning or tools.
The Geometry of Language Model Logits: Generic Structure and Learned Departures The paper analyzes the geometric distribution of language model output logits, identifying a consistent log-linear relationship that emerges during pretraining. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Theoretical study of logit rank patterns, not deployment safety.
The Geometry of Language Model Logits: Generic Structure and Learned Departures The paper analyzes the geometric distribution of language model output logits, identifying a consistent log-linear relationship that emerges during pretraining. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Analyzes pretraining logit distributions, not model compression or efficient serving.
The Geometry of Language Model Logits: Generic Structure and Learned Departures The paper analyzes the geometric distribution of language model output logits, identifying a consistent log-linear relationship that emerges during pretraining. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Investigates foundational logit geometry rather than evaluation benchmarks.

14,041–14,100 of 17,313← Previous Next →