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
Qwenfant: Train Big Character-level BabyLMs on Concise Parallel Text The paper presents an approach using character-level tokens, larger models, and synthetic data for the multilingual BabyLM data-efficiency challenge. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not address AI safety risks, robustness, or guardrails.
PersEvalBench: Can LLM Evaluators Judge Contextually Appropriate Use of Persona Information? The paper introduces a benchmark to evaluate whether language models can accurately detect appropriate and inappropriate uses of persona information in dialogue. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on benchmark evaluation of personalization rather than safety.
PersEvalBench: Can LLM Evaluators Judge Contextually Appropriate Use of Persona Information? The paper introduces a benchmark to evaluate whether language models can accurately detect appropriate and inappropriate uses of persona information in dialogue. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not address small models, distillation, or quantization.
PersEvalBench: Can LLM Evaluators Judge Contextually Appropriate Use of Persona Information? The paper introduces a benchmark to evaluate whether language models can accurately detect appropriate and inappropriate uses of persona information in dialogue. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Focuses on dialogue personalization evaluation rather than RAG.
PersEvalBench: Can LLM Evaluators Judge Contextually Appropriate Use of Persona Information? The paper introduces a benchmark to evaluate whether language models can accurately detect appropriate and inappropriate uses of persona information in dialogue. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Evaluates LLM evaluators on personalization errors, not agent planning.
PersEvalBench: Can LLM Evaluators Judge Contextually Appropriate Use of Persona Information? The paper introduces a benchmark to evaluate whether language models can accurately detect appropriate and inappropriate uses of persona information in dialogue. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses on text dialogue history rather than multimodal understanding.
PersEvalBench: Can LLM Evaluators Judge Contextually Appropriate Use of Persona Information? The paper introduces a benchmark to evaluate whether language models can accurately detect appropriate and inappropriate uses of persona information in dialogue. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Does not evaluate or generate software code.
Truthful and Expected-Payment-Optimal Model Routing in Graph-Based Agentic Workflows The paper formulates task routing across language model providers as an auction mechanism to encourage truthful cost reporting while minimizing context switching. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Addresses procurement auctions in agent workflows, not code generation.
Truthful and Expected-Payment-Optimal Model Routing in Graph-Based Agentic Workflows The paper formulates task routing across language model providers as an auction mechanism to encourage truthful cost reporting while minimizing context switching. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Studies routing economics rather than model compression or efficient serving.
Truthful and Expected-Payment-Optimal Model Routing in Graph-Based Agentic Workflows The paper formulates task routing across language model providers as an auction mechanism to encourage truthful cost reporting while minimizing context switching. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses on multi-provider agent routing without multimodal modalities.
Truthful and Expected-Payment-Optimal Model Routing in Graph-Based Agentic Workflows The paper formulates task routing across language model providers as an auction mechanism to encourage truthful cost reporting while minimizing context switching. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Focuses on provider routing mechanisms rather than RAG pipelines.
Truthful and Expected-Payment-Optimal Model Routing in Graph-Based Agentic Workflows The paper formulates task routing across language model providers as an auction mechanism to encourage truthful cost reporting while minimizing context switching. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on workflow routing algorithms rather than benchmark design.
Truthful and Expected-Payment-Optimal Model Routing in Graph-Based Agentic Workflows The paper formulates task routing across language model providers as an auction mechanism to encourage truthful cost reporting while minimizing context switching. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Addresses economic mechanism design, not AI safety or jailbreaks.
TopicMerge: A Divide and Conquer Approach to Topic Modeling The paper introduces a multi-stage topic modeling pipeline that merges fine-grained document clusters using a language model to balance efficiency and quality. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Presents a clustering pipeline rather than evaluation methodologies.
TopicMerge: A Divide and Conquer Approach to Topic Modeling The paper introduces a multi-stage topic modeling pipeline that merges fine-grained document clusters using a language model to balance efficiency and quality. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not focus on model compression, quantization, or distillation.
TopicMerge: A Divide and Conquer Approach to Topic Modeling The paper introduces a multi-stage topic modeling pipeline that merges fine-grained document clusters using a language model to balance efficiency and quality. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Deals with text document clustering without multimodal data.
TopicMerge: A Divide and Conquer Approach to Topic Modeling The paper introduces a multi-stage topic modeling pipeline that merges fine-grained document clusters using a language model to balance efficiency and quality. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not address AI safety, jailbreaks, or hallucination mitigation.
TopicMerge: A Divide and Conquer Approach to Topic Modeling The paper introduces a multi-stage topic modeling pipeline that merges fine-grained document clusters using a language model to balance efficiency and quality. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Focuses on document clustering pipelines, not multi-step agents.
TopicMerge: A Divide and Conquer Approach to Topic Modeling The paper introduces a multi-stage topic modeling pipeline that merges fine-grained document clusters using a language model to balance efficiency and quality. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Focuses on document topic modeling rather than program generation.
TopicMerge: A Divide and Conquer Approach to Topic Modeling The paper introduces a multi-stage topic modeling pipeline that merges fine-grained document clusters using a language model to balance efficiency and quality. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Proposes a topic modeling approach rather than RAG systems.
Counterfactual Gate-Support Decomposition: Separating Selection from Redistribution in Training-Data Filtering The paper proposes an experimental design that separates example selection from mass redistribution to analyze the effects of training-data quality filtering. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Analyzes training-data selection, not retrieval-augmented generation architectures.
Counterfactual Gate-Support Decomposition: Separating Selection from Redistribution in Training-Data Filtering The paper proposes an experimental design that separates example selection from mass redistribution to analyze the effects of training-data quality filtering. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Does not focus on software engineering or code generation.
Counterfactual Gate-Support Decomposition: Separating Selection from Redistribution in Training-Data Filtering The paper proposes an experimental design that separates example selection from mass redistribution to analyze the effects of training-data quality filtering. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Studies data selection effects rather than AI safety or guardrails.
Counterfactual Gate-Support Decomposition: Separating Selection from Redistribution in Training-Data Filtering The paper proposes an experimental design that separates example selection from mass redistribution to analyze the effects of training-data quality filtering. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Investigates training data filtering decomposition, not LLM agents.
Counterfactual Gate-Support Decomposition: Separating Selection from Redistribution in Training-Data Filtering The paper proposes an experimental design that separates example selection from mass redistribution to analyze the effects of training-data quality filtering. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on training data selection methodology rather than model efficiency.
Counterfactual Gate-Support Decomposition: Separating Selection from Redistribution in Training-Data Filtering The paper proposes an experimental design that separates example selection from mass redistribution to analyze the effects of training-data quality filtering. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Analyzes text retrieval data filtering without multimodal components.
Target-Token Asymmetry Can Masquerade As Depth Specialization The paper shows that differences in predicted target tokens, such as whitespace, can falsely look like layer specialization across different corpora. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Studies mechanistic interpretability across token types, not code generation.
Target-Token Asymmetry Can Masquerade As Depth Specialization The paper shows that differences in predicted target tokens, such as whitespace, can falsely look like layer specialization across different corpora. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on interpretability analysis rather than model compression techniques.
Target-Token Asymmetry Can Masquerade As Depth Specialization The paper shows that differences in predicted target tokens, such as whitespace, can falsely look like layer specialization across different corpora. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Evaluates language models on text data without multimodal elements.
Target-Token Asymmetry Can Masquerade As Depth Specialization The paper shows that differences in predicted target tokens, such as whitespace, can falsely look like layer specialization across different corpora. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Analyzes layer-ablation dynamics, not agent planning or tool use.
Target-Token Asymmetry Can Masquerade As Depth Specialization The paper shows that differences in predicted target tokens, such as whitespace, can falsely look like layer specialization across different corpora. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not address safety vulnerabilities, hallucinations, or deployment guardrails.
Target-Token Asymmetry Can Masquerade As Depth Specialization The paper shows that differences in predicted target tokens, such as whitespace, can falsely look like layer specialization across different corpora. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Does not evaluate retrieval-augmented generation systems.
Where to Start, How to Move: Analyzing Curricula under GRPO for Long-Horizon LLM Agents The paper analyzes prefix replay curricula and handoff rules under reinforcement learning to improve long-horizon task completion for language model agents. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Studies reinforcement learning curricula, not retrieval-augmented generation.
Where to Start, How to Move: Analyzing Curricula under GRPO for Long-Horizon LLM Agents The paper analyzes prefix replay curricula and handoff rules under reinforcement learning to improve long-horizon task completion for language model agents. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on training curricula under GRPO rather than evaluation methods.
Where to Start, How to Move: Analyzing Curricula under GRPO for Long-Horizon LLM Agents The paper analyzes prefix replay curricula and handoff rules under reinforcement learning to improve long-horizon task completion for language model agents. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Focuses on multi-step tool-use benchmarks rather than software engineering.
Where to Start, How to Move: Analyzing Curricula under GRPO for Long-Horizon LLM Agents The paper analyzes prefix replay curricula and handoff rules under reinforcement learning to improve long-horizon task completion for language model agents. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not address model compression, quantization, or efficient inference.
Where to Start, How to Move: Analyzing Curricula under GRPO for Long-Horizon LLM Agents The paper analyzes prefix replay curricula and handoff rules under reinforcement learning to improve long-horizon task completion for language model agents. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Studies task completion learning rather than safety or guardrails.
Where to Start, How to Move: Analyzing Curricula under GRPO for Long-Horizon LLM Agents The paper analyzes prefix replay curricula and handoff rules under reinforcement learning to improve long-horizon task completion for language model agents. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses on language-based tool calling and agent benchmarks.
VLX-Go: Robust Embodied Tracking through Multi-Stage Learning and On-Policy Refinement The paper presents a vision-language policy combining pre-aligned representations and multi-stage learning to improve robotic person tracking among distractors. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on robotic control rather than model safety or alignment.
VLX-Go: Robust Embodied Tracking through Multi-Stage Learning and On-Policy Refinement The paper presents a vision-language policy combining pre-aligned representations and multi-stage learning to improve robotic person tracking among distractors. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Focuses on robotic visual tracking, not programming code.
VLX-Go: Robust Embodied Tracking through Multi-Stage Learning and On-Policy Refinement The paper presents a vision-language policy combining pre-aligned representations and multi-stage learning to improve robotic person tracking among distractors. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes an embodied tracking policy rather than evaluation frameworks.
VLX-Go: Robust Embodied Tracking through Multi-Stage Learning and On-Policy Refinement The paper presents a vision-language policy combining pre-aligned representations and multi-stage learning to improve robotic person tracking among distractors. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not focus on model compression or inference acceleration.
VLX-Go: Robust Embodied Tracking through Multi-Stage Learning and On-Policy Refinement The paper presents a vision-language policy combining pre-aligned representations and multi-stage learning to improve robotic person tracking among distractors. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 15:06	Not relevant Studies embodied robot person-following, not industrial inspection.
VLX-Go: Robust Embodied Tracking through Multi-Stage Learning and On-Policy Refinement The paper presents a vision-language policy combining pre-aligned representations and multi-stage learning to improve robotic person tracking among distractors. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Does not investigate retrieval-augmented text generation methods.
Learning What to Keep: Transfer-Aware Personal Skill Learning from User Feedback The paper introduces a framework that decomposes user feedback into atomic rules to retain only reusable preferences for personalizing language models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Studies preference learning rather than safety, guardrails, or jailbreaks.
Learning What to Keep: Transfer-Aware Personal Skill Learning from User Feedback The paper introduces a framework that decomposes user feedback into atomic rules to retain only reusable preferences for personalizing language models. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Focuses on personalized skill induction rather than RAG systems.
Learning What to Keep: Transfer-Aware Personal Skill Learning from User Feedback The paper introduces a framework that decomposes user feedback into atomic rules to retain only reusable preferences for personalizing language models. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not investigate pruning, distillation, or lightweight models.
Learning What to Keep: Transfer-Aware Personal Skill Learning from User Feedback The paper introduces a framework that decomposes user feedback into atomic rules to retain only reusable preferences for personalizing language models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Focuses on skill acquisition methods rather than benchmark reliability.
Learning What to Keep: Transfer-Aware Personal Skill Learning from User Feedback The paper introduces a framework that decomposes user feedback into atomic rules to retain only reusable preferences for personalizing language models. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Deals purely with textual user feedback and personalization.
Learning What to Keep: Transfer-Aware Personal Skill Learning from User Feedback The paper introduces a framework that decomposes user feedback into atomic rules to retain only reusable preferences for personalizing language models. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Evaluates personalized writing tasks rather than software engineering.
The Sum of All Summaries: PROPALIGN for Graded Proposition Salience Evaluation The paper proposes a multiple-summary metric and dataset to evaluate graded proposition salience across diverse text genres. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Does not focus on model compression or efficient serving.
The Sum of All Summaries: PROPALIGN for Graded Proposition Salience Evaluation The paper proposes a multiple-summary metric and dataset to evaluate graded proposition salience across diverse text genres. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Deals with written and spoken English text, not multimodal.
The Sum of All Summaries: PROPALIGN for Graded Proposition Salience Evaluation The paper proposes a multiple-summary metric and dataset to evaluate graded proposition salience across diverse text genres. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Does not evaluate external document retrieval or grounding architectures.
The Sum of All Summaries: PROPALIGN for Graded Proposition Salience Evaluation The paper proposes a multiple-summary metric and dataset to evaluate graded proposition salience across diverse text genres. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Does not focus on safety vulnerabilities or deployment reliability.
The Sum of All Summaries: PROPALIGN for Graded Proposition Salience Evaluation The paper proposes a multiple-summary metric and dataset to evaluate graded proposition salience across diverse text genres. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Applies to general text summarization rather than code generation.
The Sum of All Summaries: PROPALIGN for Graded Proposition Salience Evaluation The paper proposes a multiple-summary metric and dataset to evaluate graded proposition salience across diverse text genres. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Investigates proposition salience extraction, not multi-step agent planning.
ForkGRPO: Efficient Exploration and Fine-Grained Credit Assignment for Diffusion Language Models The paper introduces a reinforcement learning framework using intermediate-state branching and segment-level credit assignment for diffusion language models. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Studies exploration in diffusion language models, not model efficiency.
ForkGRPO: Efficient Exploration and Fine-Grained Credit Assignment for Diffusion Language Models The paper introduces a reinforcement learning framework using intermediate-state branching and segment-level credit assignment for diffusion language models. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Studies reinforcement learning for diffusion language models, not agent workflows.
ForkGRPO: Efficient Exploration and Fine-Grained Credit Assignment for Diffusion Language Models The paper introduces a reinforcement learning framework using intermediate-state branching and segment-level credit assignment for diffusion language models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on reinforcement learning optimization, not safety guardrails.
ForkGRPO: Efficient Exploration and Fine-Grained Credit Assignment for Diffusion Language Models The paper introduces a reinforcement learning framework using intermediate-state branching and segment-level credit assignment for diffusion language models. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses on diffusion language models without multimodal inputs.

6,721–6,780 of 17,313← Previous Next →