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
Understanding Machine Taste: Aesthetic Agreement Across Foundation Models in K-pop The paper analyzes aesthetic judgments by foundation models on K-pop music videos to measure inter-model consensus across audio and visual modalities. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Studies creative aesthetic ranking rather than safety, robustness, or guardrails.
Understanding Machine Taste: Aesthetic Agreement Across Foundation Models in K-pop The paper analyzes aesthetic judgments by foundation models on K-pop music videos to measure inter-model consensus across audio and visual modalities. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies aesthetic preference ranking rather than multi-step agent systems.
Understanding Machine Taste: Aesthetic Agreement Across Foundation Models in K-pop The paper analyzes aesthetic judgments by foundation models on K-pop music videos to measure inter-model consensus across audio and visual modalities. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on music and video aesthetics, not software engineering.
Understanding Machine Taste: Aesthetic Agreement Across Foundation Models in K-pop The paper analyzes aesthetic judgments by foundation models on K-pop music videos to measure inter-model consensus across audio and visual modalities. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Does not touch upon retrieval-augmented generation or document search.
Understanding Machine Taste: Aesthetic Agreement Across Foundation Models in K-pop The paper analyzes aesthetic judgments by foundation models on K-pop music videos to measure inter-model consensus across audio and visual modalities. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Does not address model compression, pruning, or serving speed.
Geometric Structure Predicts Per-Example Neighborhood Loss Under PCA Truncation The paper analyzes geometric properties of text embeddings to predict how many nearest neighbors an individual example loses under dimensionality reduction. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Focuses on text embeddings without image, video, or audio components.
Geometric Structure Predicts Per-Example Neighborhood Loss Under PCA Truncation The paper analyzes geometric properties of text embeddings to predict how many nearest neighbors an individual example loses under dimensionality reduction. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Analyzes embedding dimension reduction rather than autonomous agents.
Geometric Structure Predicts Per-Example Neighborhood Loss Under PCA Truncation The paper analyzes geometric properties of text embeddings to predict how many nearest neighbors an individual example loses under dimensionality reduction. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on embedding geometry rather than code generation or repair.
Geometric Structure Predicts Per-Example Neighborhood Loss Under PCA Truncation The paper analyzes geometric properties of text embeddings to predict how many nearest neighbors an individual example loses under dimensionality reduction. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Addresses embedding geometry and compression rather than safety, alignment, or guardrails.
Geometric Structure Predicts Per-Example Neighborhood Loss Under PCA Truncation The paper analyzes geometric properties of text embeddings to predict how many nearest neighbors an individual example loses under dimensionality reduction. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Analyzes geometric stability under PCA truncation rather than establishing LLM benchmarks.
Connecting the Dots: Enduring Performance Gaps in ASR Systems Between Standard and Minority English Varieties The paper tracks automatic speech recognition performance from 2021 to 2025, finding persistent accuracy disparities between standard American English and minority accents. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Investigates speech recognition systems, not software engineering or code.
Connecting the Dots: Enduring Performance Gaps in ASR Systems Between Standard and Minority English Varieties The paper tracks automatic speech recognition performance from 2021 to 2025, finding persistent accuracy disparities between standard American English and minority accents. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Focuses on accent fairness rather than model compression or inference efficiency.
Connecting the Dots: Enduring Performance Gaps in ASR Systems Between Standard and Minority English Varieties The paper tracks automatic speech recognition performance from 2021 to 2025, finding persistent accuracy disparities between standard American English and minority accents. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on speech recognition rather than retrieval-augmented generation.
Connecting the Dots: Enduring Performance Gaps in ASR Systems Between Standard and Minority English Varieties The paper tracks automatic speech recognition performance from 2021 to 2025, finding persistent accuracy disparities between standard American English and minority accents. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Evaluates speech recognition models rather than autonomous agents.
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. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Analyzes text sentence highlighting without visual or audio modalities.
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. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on social highlight prediction rather than retrieval-augmented generation.
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. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on user preference modeling rather than safety, alignment, or guardrails.
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. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies highlight prediction models rather than autonomous multi-step agents.
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. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Does not relate to software engineering or code LLMs.
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. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Compares training data scaling versus model teachers rather than proposing LLM benchmarks.
Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders The paper proposes ReMaRk, a reinforcement learning method that trains recommender models using all available reader replies simultaneously to reduce variance. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses recommendation systems, not program repair or code generation.
Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders The paper proposes ReMaRk, a reinforcement learning method that trains recommender models using all available reader replies simultaneously to reduce variance. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Deals purely with text recommendation threads without multimodal components.
Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders The paper proposes ReMaRk, a reinforcement learning method that trains recommender models using all available reader replies simultaneously to reduce variance. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on training efficiency for recommenders rather than safety guardrails or hallucination.
Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders The paper proposes ReMaRk, a reinforcement learning method that trains recommender models using all available reader replies simultaneously to reduce variance. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on RL training recipes for recommenders rather than autonomous agents.
Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders The paper proposes ReMaRk, a reinforcement learning method that trains recommender models using all available reader replies simultaneously to reduce variance. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Deals with recommendation reward functions rather than retrieval-augmented generation.
Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders The paper proposes ReMaRk, a reinforcement learning method that trains recommender models using all available reader replies simultaneously to reduce variance. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Focuses on RL training optimization rather than evaluation methodology or benchmarks.
Beyond Belief Nesting: Rethinking Higher-Order Theory of Mind in Large Language Models The paper introduces HopToM, a benchmark testing whether higher-order theory-of-mind tasks in language models genuinely evaluate nested reasoning rather than shortcuts. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Studies perspective reasoning capabilities rather than safety risks, jailbreaks, or guardrails.
Beyond Belief Nesting: Rethinking Higher-Order Theory of Mind in Large Language Models The paper introduces HopToM, a benchmark testing whether higher-order theory-of-mind tasks in language models genuinely evaluate nested reasoning rather than shortcuts. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Examines narrative Theory of Mind reasoning, not code generation.
Beyond Belief Nesting: Rethinking Higher-Order Theory of Mind in Large Language Models The paper introduces HopToM, a benchmark testing whether higher-order theory-of-mind tasks in language models genuinely evaluate nested reasoning rather than shortcuts. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Evaluates text-based reasoning without multimodal data.
Beyond Belief Nesting: Rethinking Higher-Order Theory of Mind in Large Language Models The paper introduces HopToM, a benchmark testing whether higher-order theory-of-mind tasks in language models genuinely evaluate nested reasoning rather than shortcuts. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Does not cover external knowledge retrieval or grounding.
Beyond Belief Nesting: Rethinking Higher-Order Theory of Mind in Large Language Models The paper introduces HopToM, a benchmark testing whether higher-order theory-of-mind tasks in language models genuinely evaluate nested reasoning rather than shortcuts. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on Theory of Mind reasoning rather than tool use or autonomous agents.
Beyond Belief Nesting: Rethinking Higher-Order Theory of Mind in Large Language Models The paper introduces HopToM, a benchmark testing whether higher-order theory-of-mind tasks in language models genuinely evaluate nested reasoning rather than shortcuts. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Focuses on reasoning accuracy rather than model compression or inference efficiency.
GENDIS: Gender-Related Discourse and Stance in Multilingual YouTube Comments from India, Kenya, and Nigeria The paper presents GENDIS, a multilingual dataset of YouTube comments from India, Kenya, and Nigeria annotated for gender-related discourse and stance. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Deals with social media comments, not programming or code assistance.
GENDIS: Gender-Related Discourse and Stance in Multilingual YouTube Comments from India, Kenya, and Nigeria The paper presents GENDIS, a multilingual dataset of YouTube comments from India, Kenya, and Nigeria annotated for gender-related discourse and stance. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on comment classification rather than retrieval-augmented generation.
GENDIS: Gender-Related Discourse and Stance in Multilingual YouTube Comments from India, Kenya, and Nigeria The paper presents GENDIS, a multilingual dataset of YouTube comments from India, Kenya, and Nigeria annotated for gender-related discourse and stance. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Does not focus on model compression, quantization, or efficient architectures.
GENDIS: Gender-Related Discourse and Stance in Multilingual YouTube Comments from India, Kenya, and Nigeria The paper presents GENDIS, a multilingual dataset of YouTube comments from India, Kenya, and Nigeria annotated for gender-related discourse and stance. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Presents a discourse dataset without agent planning or tool use.
GENDIS: Gender-Related Discourse and Stance in Multilingual YouTube Comments from India, Kenya, and Nigeria The paper presents GENDIS, a multilingual dataset of YouTube comments from India, Kenya, and Nigeria annotated for gender-related discourse and stance. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Analyzes text comments from YouTube without multimodal image or audio models.
Full-duplex Spoken Language Modeling with Discrete Diffusion The paper presents Diffuplex, a full-duplex spoken language model that separates speech timing from content generation using discrete diffusion. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Does not involve external document retrieval or grounding.
Full-duplex Spoken Language Modeling with Discrete Diffusion The paper presents Diffuplex, a full-duplex spoken language model that separates speech timing from content generation using discrete diffusion. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on spoken dialogue modeling rather than code generation.
Full-duplex Spoken Language Modeling with Discrete Diffusion The paper presents Diffuplex, a full-duplex spoken language model that separates speech timing from content generation using discrete diffusion. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Presents a spoken language model rather than an evaluation benchmark.
Full-duplex Spoken Language Modeling with Discrete Diffusion The paper presents Diffuplex, a full-duplex spoken language model that separates speech timing from content generation using discrete diffusion. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Proposes a speech model architecture rather than an autonomous tool-using agent.
Full-duplex Spoken Language Modeling with Discrete Diffusion The paper presents Diffuplex, a full-duplex spoken language model that separates speech timing from content generation using discrete diffusion. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on latency and spoken dialogue architecture rather than safety or guardrails.
Where Words Begin and End The paper evaluates audio forced aligners and fine-tunes Whisper to predict human-aligned word boundaries across read, conversational, and telephone speech. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Addresses acoustic-text alignment, not retrieval-augmented generation.
Where Words Begin and End The paper evaluates audio forced aligners and fine-tunes Whisper to predict human-aligned word boundaries across read, conversational, and telephone speech. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Focuses on fine-tuning speech models rather than general LLM evaluation benchmarks.
Where Words Begin and End The paper evaluates audio forced aligners and fine-tunes Whisper to predict human-aligned word boundaries across read, conversational, and telephone speech. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Does not focus on model compression, quantization, or serving efficiency.
Where Words Begin and End The paper evaluates audio forced aligners and fine-tunes Whisper to predict human-aligned word boundaries across read, conversational, and telephone speech. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on forced speech alignment algorithms rather than agent workflows.
Where Words Begin and End The paper evaluates audio forced aligners and fine-tunes Whisper to predict human-aligned word boundaries across read, conversational, and telephone speech. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Investigates word boundary timing accuracy rather than safety or reliability guardrails.
Where Words Begin and End The paper evaluates audio forced aligners and fine-tunes Whisper to predict human-aligned word boundaries across read, conversational, and telephone speech. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Deals with speech-to-text alignment, not software engineering or code.
BRIDGE: Selective Boundary Repair with Dependency Graph Evidence in Multi-Agent Systems The paper introduces BRIDGE, a runtime intervention framework that detects and repairs broken information handoffs across dependency graphs in multi-agent workflows. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses general multi-agent task collaboration rather than software engineering tasks.
BRIDGE: Selective Boundary Repair with Dependency Graph Evidence in Multi-Agent Systems The paper introduces BRIDGE, a runtime intervention framework that detects and repairs broken information handoffs across dependency graphs in multi-agent workflows. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Does not address parameter reduction, distillation, or inference optimization.
BRIDGE: Selective Boundary Repair with Dependency Graph Evidence in Multi-Agent Systems The paper introduces BRIDGE, a runtime intervention framework that detects and repairs broken information handoffs across dependency graphs in multi-agent workflows. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Proposes a multi-agent repair framework rather than an evaluation benchmark.
BRIDGE: Selective Boundary Repair with Dependency Graph Evidence in Multi-Agent Systems The paper introduces BRIDGE, a runtime intervention framework that detects and repairs broken information handoffs across dependency graphs in multi-agent workflows. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Investigates multi-agent interaction traces without multimodal vision or audio data.
BRIDGE: Selective Boundary Repair with Dependency Graph Evidence in Multi-Agent Systems The paper introduces BRIDGE, a runtime intervention framework that detects and repairs broken information handoffs across dependency graphs in multi-agent workflows. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on multi-agent communication handoffs rather than external knowledge retrieval.
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. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on watermark robustness without information retrieval components.
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. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Focuses on watermarking security rather than model efficiency or compression.
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. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Investigates text watermarking techniques rather than multi-step agent systems.
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. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Deals exclusively with text representations and generation.
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. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Evaluates watermark robustness rather than general LLM capability benchmarks.
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. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses general text watermarking, not code generation or program repair.
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. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on architectural efficiency rather than deployment safety or jailbreak prevention.

11,341–11,400 of 17,313← Previous Next →