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
Gradient Routing Localizes and Removes Unintended Behaviors in RL Routes gradients from imperfect monitor flags into sacrificial parameters to mitigate reward hacking during reinforcement learning. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Presents an optimization technique rather than an evaluation benchmark.
Superintelligence and the Triple Risk of an AI Race Argues that competitive national races toward superintelligent artificial intelligence create risks of conflict and catastrophic loss of control. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant A strategic policy discussion rather than model benchmarking or testing.
Superintelligence and the Triple Risk of an AI Race Argues that competitive national races toward superintelligent artificial intelligence create risks of conflict and catastrophic loss of control. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Policy essay on geopolitical competition, not agent technical architectures.
Interactive Navigation Beyond Reachability: Benchmark and Agentic Framework Introduces a benchmark and agent framework for embodied navigation tasks requiring physical environment interactions like opening doors. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on navigation capabilities rather than safety risks or guardrails.
Interactive Navigation Beyond Reachability: Benchmark and Agentic Framework Introduces a benchmark and agent framework for embodied navigation tasks requiring physical environment interactions like opening doors. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Embodied indoor navigation, not manufacturing defect detection or industrial quality control.
What the Machine Invented: Mining New Mathematical Instruments from OpenAI's October 2026 Release of AI-Generated Mathematics Extracts and categorizes mathematical proof mechanisms and lemmas from an open release of synthetic mathematical manuscripts. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Discusses mathematics generation rather than AI deployment risks or safety.
What the Machine Invented: Mining New Mathematical Instruments from OpenAI's October 2026 Release of AI-Generated Mathematics Extracts and categorizes mathematical proof mechanisms and lemmas from an open release of synthetic mathematical manuscripts. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Analyzes AI-generated mathematical manuscripts rather than evaluating model benchmarks.
What the Machine Invented: Mining New Mathematical Instruments from OpenAI's October 2026 Release of AI-Generated Mathematics Extracts and categorizes mathematical proof mechanisms and lemmas from an open release of synthetic mathematical manuscripts. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies mathematical proofs rather than autonomous tool-using agents.
SFT Leaves a Reproducible Signature in Activation Geometry: Calibrating Post-Training Interpretability Against Run-to-Run Noise Measures post-training activation geometry shifts against run-to-run noise, identifying distinct structural signatures left by supervised fine-tuning. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Studies activation geometry variance rather than safety guardrails or red-teaming.
SFT Leaves a Reproducible Signature in Activation Geometry: Calibrating Post-Training Interpretability Against Run-to-Run Noise Measures post-training activation geometry shifts against run-to-run noise, identifying distinct structural signatures left by supervised fine-tuning. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Examines neural representations rather than multi-step autonomous planning agents.
AIPOCH Open-Science: An Open-Source AI Research Workbench with Reproducible .science Packages Presents an open-source research workbench and portable packaging format to preserve relationships between computations, artifacts, and assessments. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant About scientific workflow reproducibility rather than software engineering code generation.
AIPOCH Open-Science: An Open-Source AI Research Workbench with Reproducible .science Packages Presents an open-source research workbench and portable packaging format to preserve relationships between computations, artifacts, and assessments. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on research workflow software rather than autonomous planning agents.
AIPOCH Open-Science: An Open-Source AI Research Workbench with Reproducible .science Packages Presents an open-source research workbench and portable packaging format to preserve relationships between computations, artifacts, and assessments. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Covers scientific reproducibility packages rather than AI safety or deployment guardrails.
AIPOCH Open-Science: An Open-Source AI Research Workbench with Reproducible .science Packages Presents an open-source research workbench and portable packaging format to preserve relationships between computations, artifacts, and assessments. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Introduces a reproducible workbench rather than an LLM evaluation benchmark.
Calibrated Typed Decisions for Entity, Relation, and Event Extraction Presents a unified encoder that jointly models type definitions and text for cross-domain entity, relation, and event extraction. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Focuses on text extraction rather than multimodal or vision-language tasks.
Calibrated Typed Decisions for Entity, Relation, and Event Extraction Presents a unified encoder that jointly models type definitions and text for cross-domain entity, relation, and event extraction. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Addresses entity and relation extraction, not retrieval-augmented generation architectures.
Calibrated Typed Decisions for Entity, Relation, and Event Extraction Presents a unified encoder that jointly models type definitions and text for cross-domain entity, relation, and event extraction. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on extraction accuracy rather than deployment safety, hallucinations, or jailbreaks.
Calibrated Typed Decisions for Entity, Relation, and Event Extraction Presents a unified encoder that jointly models type definitions and text for cross-domain entity, relation, and event extraction. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on information extraction rather than multi-step tool-using agents.
Calibrated Typed Decisions for Entity, Relation, and Event Extraction Presents a unified encoder that jointly models type definitions and text for cross-domain entity, relation, and event extraction. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant About NLP information extraction rather than code synthesis or software engineering.
Calibrated Typed Decisions for Entity, Relation, and Event Extraction Presents a unified encoder that jointly models type definitions and text for cross-domain entity, relation, and event extraction. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Evaluates on benchmarks but primarily introduces a new extraction model architecture.
D-World: An Explicit World-Language-Action Model with Recursive Self-Improvement under Dual Control Designs an explicit digital-twin world model where robotic agents inspect, simulate, and calibrate physical parameters before acting. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Proposes a model framework rather than evaluating benchmarks or testing methodologies.
D-World: An Explicit World-Language-Action Model with Recursive Self-Improvement under Dual Control Designs an explicit digital-twin world model where robotic agents inspect, simulate, and calibrate physical parameters before acting. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on robotics and control rather than AI safety guardrails or hallucination.
D-World: An Explicit World-Language-Action Model with Recursive Self-Improvement under Dual Control Designs an explicit digital-twin world model where robotic agents inspect, simulate, and calibrate physical parameters before acting. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Focuses on embodied robotic control, not industrial inspection or defect detection.
When LLM Agents See a Conflict but Still Proceed: Cross-Domain Dependencies in Network Slicing Shows that language agents often ignore identified cross-domain dependencies, causing unintended disruptive actions in network management. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Evaluates agent action execution, not retrieval-augmented generation systems.
When LLM Agents See a Conflict but Still Proceed: Cross-Domain Dependencies in Network Slicing Shows that language agents often ignore identified cross-domain dependencies, causing unintended disruptive actions in network management. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Examines text-based network management intents, not multimodal models.
When LLM Agents See a Conflict but Still Proceed: Cross-Domain Dependencies in Network Slicing Shows that language agents often ignore identified cross-domain dependencies, causing unintended disruptive actions in network management. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on network intent operations, not code generation assistants.
When LLM Agents See a Conflict but Still Proceed: Cross-Domain Dependencies in Network Slicing Shows that language agents often ignore identified cross-domain dependencies, causing unintended disruptive actions in network management. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Focuses on agent decision reliability, not efficient inference techniques.
D2: High-Capability Full-Duplex Spoken Language Models from Half-Duplex Backbones Converts half-duplex speech models into streaming full-duplex conversational models capable of simultaneous listening and speaking. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Spoken dialogue modeling, not code generation or program repair.
D2: High-Capability Full-Duplex Spoken Language Models from Half-Duplex Backbones Converts half-duplex speech models into streaming full-duplex conversational models capable of simultaneous listening and speaking. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Covers spoken turn-taking mechanics, not deployment safety guardrails.
D2: High-Capability Full-Duplex Spoken Language Models from Half-Duplex Backbones Converts half-duplex speech models into streaming full-duplex conversational models capable of simultaneous listening and speaking. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Addresses full-duplex conversion, not model pruning or quantization.
D2: High-Capability Full-Duplex Spoken Language Models from Half-Duplex Backbones Converts half-duplex speech models into streaming full-duplex conversational models capable of simultaneous listening and speaking. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on duplex speech modeling, not multi-step agent planning.
D2: High-Capability Full-Duplex Spoken Language Models from Half-Duplex Backbones Converts half-duplex speech models into streaming full-duplex conversational models capable of simultaneous listening and speaking. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Model conversion method, not an evaluation benchmark or audit.
D2: High-Capability Full-Duplex Spoken Language Models from Half-Duplex Backbones Converts half-duplex speech models into streaming full-duplex conversational models capable of simultaneous listening and speaking. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Spoken interaction modeling, not retrieval-augmented generation.
Who is the 'I' in AI? Establishing Decentralized Identity and Accountability for Autonomous Agents Proposes a decentralized identity framework using self-sovereign identity principles to establish verification and accountability for autonomous agents. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Identity specification proposal, not an evaluation benchmark.
EvoQuantLab: Training-Free Dual-Loop Evolution for Auditable Quantitative Research Combines genetic programming and diagnostic evaluation loops to discover quantitative financial trading factors without backtest overfitting. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Proposes a factor discovery system, not an LLM benchmark.
EvoQuantLab: Training-Free Dual-Loop Evolution for Auditable Quantitative Research Combines genetic programming and diagnostic evaluation loops to discover quantitative financial trading factors without backtest overfitting. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Financial factor mining methodology, not AI deployment safety.
SonicWorld: Source-Centric Latent Dynamics for Spatial Audio Reasoning Introduces a dynamic spatial audio reasoning benchmark and an architecture tracking acoustic sources across moving indoor scenes. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on spatial audio understanding, not safety guardrails.
SonicWorld: Source-Centric Latent Dynamics for Spatial Audio Reasoning Introduces a dynamic spatial audio reasoning benchmark and an architecture tracking acoustic sources across moving indoor scenes. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies spatial audio reasoning models, not autonomous LLM agents.
The Format is the Filter: Scoring Format Fixes What an LLM Judge Can Express Demonstrates that the scoring format chosen for an evaluation model strongly restricts its expressive capacity and ranking reliability. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on judging instruction data, not document retrieval systems.
The Format is the Filter: Scoring Format Fixes What an LLM Judge Can Express Demonstrates that the scoring format chosen for an evaluation model strongly restricts its expressive capacity and ranking reliability. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Evaluates general instruction-tuning curation, not coding assistance.
The Format is the Filter: Scoring Format Fixes What an LLM Judge Can Express Demonstrates that the scoring format chosen for an evaluation model strongly restricts its expressive capacity and ranking reliability. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies LLM-as-a-judge scoring formats, not multi-step agent workflows.
The Format is the Filter: Scoring Format Fixes What an LLM Judge Can Express Demonstrates that the scoring format chosen for an evaluation model strongly restricts its expressive capacity and ranking reliability. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Data curation format study, not model compression or efficiency.
The Format is the Filter: Scoring Format Fixes What an LLM Judge Can Express Demonstrates that the scoring format chosen for an evaluation model strongly restricts its expressive capacity and ranking reliability. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Studies evaluation formatting artifacts, not deployment guardrails or jailbreaks.
The Format is the Filter: Scoring Format Fixes What an LLM Judge Can Express Demonstrates that the scoring format chosen for an evaluation model strongly restricts its expressive capacity and ranking reliability. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Evaluates text scoring formats, not multimodal models.
Force-Modulated Dexterity: Unified Compliance for Contact-Rich Manipulation Presents a compliant teleoperation control framework that reduces excessive contact forces during dexterous manipulation data collection. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Covers robot teleoperation compliance, not AI deployment guardrails.
Force-Modulated Dexterity: Unified Compliance for Contact-Rich Manipulation Presents a compliant teleoperation control framework that reduces excessive contact forces during dexterous manipulation data collection. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Proposes a robotic control technique, not LLM evaluation.
Force-Modulated Dexterity: Unified Compliance for Contact-Rich Manipulation Presents a compliant teleoperation control framework that reduces excessive contact forces during dexterous manipulation data collection. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on teleoperated robotics control, not autonomous LLM agents.
Force-Modulated Dexterity: Unified Compliance for Contact-Rich Manipulation Presents a compliant teleoperation control framework that reduces excessive contact forces during dexterous manipulation data collection. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Robotic teleoperation and compliance control, not industrial vision inspection.
Learning Perceptual Geometry of Natural Odor Mixtures from LLM-Derived Descriptors Uses language-model perceptual descriptions of complex chemical mixtures to predict perceived odor similarity from molecular composition. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Maps chemical mixtures to sensory space, not vision/speech multimodality.
Learning Perceptual Geometry of Natural Odor Mixtures from LLM-Derived Descriptors Uses language-model perceptual descriptions of complex chemical mixtures to predict perceived odor similarity from molecular composition. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on molecular property prediction, not retrieval-augmented generation.
Learning Perceptual Geometry of Natural Odor Mixtures from LLM-Derived Descriptors Uses language-model perceptual descriptions of complex chemical mixtures to predict perceived odor similarity from molecular composition. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Study on odor perception representations, not LLM benchmark evaluation.
Learning Perceptual Geometry of Natural Odor Mixtures from LLM-Derived Descriptors Uses language-model perceptual descriptions of complex chemical mixtures to predict perceived odor similarity from molecular composition. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on odor geometry, not AI deployment safety.
Learning Perceptual Geometry of Natural Odor Mixtures from LLM-Derived Descriptors Uses language-model perceptual descriptions of complex chemical mixtures to predict perceived odor similarity from molecular composition. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Application in computational chemistry and olfaction, not code generation.
Learning Perceptual Geometry of Natural Odor Mixtures from LLM-Derived Descriptors Uses language-model perceptual descriptions of complex chemical mixtures to predict perceived odor similarity from molecular composition. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Uses LLM embeddings for chemical biology, not autonomous agents.
Learning Perceptual Geometry of Natural Odor Mixtures from LLM-Derived Descriptors Uses language-model perceptual descriptions of complex chemical mixtures to predict perceived odor similarity from molecular composition. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Scientific olfaction study, not efficient serving or model quantization.
Generative Flow on Symbolic Abstractions for Embodiment Capability Discovery Discovers robot capabilities using a generative flow network that creates symbolic task skeletons grounded in physical robot affordances. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on robot task learning, not AI safety guardrails.
Generative Flow on Symbolic Abstractions for Embodiment Capability Discovery Discovers robot capabilities using a generative flow network that creates symbolic task skeletons grounded in physical robot affordances. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Proposes a task generation framework, not an evaluation benchmark.
Generative Flow on Symbolic Abstractions for Embodiment Capability Discovery Discovers robot capabilities using a generative flow network that creates symbolic task skeletons grounded in physical robot affordances. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Robotics curriculum discovery using GFlowNets, not language agent systems.
Generative Flow on Symbolic Abstractions for Embodiment Capability Discovery Discovers robot capabilities using a generative flow network that creates symbolic task skeletons grounded in physical robot affordances. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant General robot manipulation discovery, not industrial defect detection.
Generative Flow on Symbolic Abstractions for Embodiment Capability Discovery Discovers robot capabilities using a generative flow network that creates symbolic task skeletons grounded in physical robot affordances. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Focuses on robotic skill acquisition, not model compression.

13,201–13,260 of 17,313← Previous Next →