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
Support or Magnitude? Characterizing Neuron Relevance for Instruction Data Selection This paper examines neuron relevance characterization methods based on magnitude and distribution support for selecting instruction fine-tuning data. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:18	Not relevant Studies training data selection rather than safety guardrails, hallucinations, or attacks.
Support or Magnitude? Characterizing Neuron Relevance for Instruction Data Selection This paper examines neuron relevance characterization methods based on magnitude and distribution support for selecting instruction fine-tuning data. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18	Not relevant Proposes data curation metrics rather than benchmarks or evaluation frameworks.
Spectral-Driven Adaptive-Rank Optimizer for Large Language Model Reinforcement Learning with Verifiable Rewards This paper investigates gradient spectral properties during reinforcement learning with verifiable rewards to design an adaptive-rank optimizer for reasoning models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Optimizes training dynamics for reasoning rather than assessing safety or reliability vulnerabilities.
Spectral-Driven Adaptive-Rank Optimizer for Large Language Model Reinforcement Learning with Verifiable Rewards This paper investigates gradient spectral properties during reinforcement learning with verifiable rewards to design an adaptive-rank optimizer for reasoning models. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Proposes a training optimizer for reinforcement learning, not agent tool execution.
Spectral-Driven Adaptive-Rank Optimizer for Large Language Model Reinforcement Learning with Verifiable Rewards This paper investigates gradient spectral properties during reinforcement learning with verifiable rewards to design an adaptive-rank optimizer for reasoning models. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Presents a general optimization algorithm for LLM reasoning, not software engineering.
Spectral-Driven Adaptive-Rank Optimizer for Large Language Model Reinforcement Learning with Verifiable Rewards This paper investigates gradient spectral properties during reinforcement learning with verifiable rewards to design an adaptive-rank optimizer for reasoning models. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Focuses on optimization methods rather than benchmarking or evaluation protocols.
Spectral-Driven Adaptive-Rank Optimizer for Large Language Model Reinforcement Learning with Verifiable Rewards This paper investigates gradient spectral properties during reinforcement learning with verifiable rewards to design an adaptive-rank optimizer for reasoning models. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Focuses on language model training optimization without multimodal data.
Spectral-Driven Adaptive-Rank Optimizer for Large Language Model Reinforcement Learning with Verifiable Rewards This paper investigates gradient spectral properties during reinforcement learning with verifiable rewards to design an adaptive-rank optimizer for reasoning models. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Studies optimizer geometry in RLVR rather than retrieval or document grounding.
Know Yourself and Know Your Neighbour: A Syntactically Informed Self-Supervised Compositional Sentence Representation Learning Framework using a Recursive Hypernetwork This paper presents a self-supervised recursive hypernetwork framework using parse trees to learn syntactically informed compositional sentence representations. Peer-reviewed	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Addresses representation learning without evaluating deployment safety, alignment, or guardrails.
Know Yourself and Know Your Neighbour: A Syntactically Informed Self-Supervised Compositional Sentence Representation Learning Framework using a Recursive Hypernetwork This paper presents a self-supervised recursive hypernetwork framework using parse trees to learn syntactically informed compositional sentence representations. Peer-reviewed	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Evaluates representations on probing tasks rather than providing an LLM evaluation benchmark.
Know Yourself and Know Your Neighbour: A Syntactically Informed Self-Supervised Compositional Sentence Representation Learning Framework using a Recursive Hypernetwork This paper presents a self-supervised recursive hypernetwork framework using parse trees to learn syntactically informed compositional sentence representations. Peer-reviewed	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Proposes a recursive hypernetwork architecture without addressing compression or serving efficiency.
Know Yourself and Know Your Neighbour: A Syntactically Informed Self-Supervised Compositional Sentence Representation Learning Framework using a Recursive Hypernetwork This paper presents a self-supervised recursive hypernetwork framework using parse trees to learn syntactically informed compositional sentence representations. Peer-reviewed	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on compositional sentence embeddings, not agent planning or tool execution.
Know Yourself and Know Your Neighbour: A Syntactically Informed Self-Supervised Compositional Sentence Representation Learning Framework using a Recursive Hypernetwork This paper presents a self-supervised recursive hypernetwork framework using parse trees to learn syntactically informed compositional sentence representations. Peer-reviewed	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Works strictly on text sentence representations without any multimodal features.
Know Yourself and Know Your Neighbour: A Syntactically Informed Self-Supervised Compositional Sentence Representation Learning Framework using a Recursive Hypernetwork This paper presents a self-supervised recursive hypernetwork framework using parse trees to learn syntactically informed compositional sentence representations. Peer-reviewed	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Proposes sentence representation learning rather than retrieval-augmented generation systems.
Know Yourself and Know Your Neighbour: A Syntactically Informed Self-Supervised Compositional Sentence Representation Learning Framework using a Recursive Hypernetwork This paper presents a self-supervised recursive hypernetwork framework using parse trees to learn syntactically informed compositional sentence representations. Peer-reviewed	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Studies natural language sentence syntax and semantics, not code generation.
SIREN: Situation-aware Index for Response and Emergency ActioNability in Crisis Translation This paper introduces an evaluation index to measure whether machine translation of emergency communication preserves critical safety actionability. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on crisis translation metrics, not autonomous planning or tool-using agents.
SIREN: Situation-aware Index for Response and Emergency ActioNability in Crisis Translation This paper introduces an evaluation index to measure whether machine translation of emergency communication preserves critical safety actionability. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Assesses translation quality and impact rather than model compression or inference efficiency.
SIREN: Situation-aware Index for Response and Emergency ActioNability in Crisis Translation This paper introduces an evaluation index to measure whether machine translation of emergency communication preserves critical safety actionability. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Addresses machine translation evaluation rather than retrieval-augmented generation systems.
SIREN: Situation-aware Index for Response and Emergency ActioNability in Crisis Translation This paper introduces an evaluation index to measure whether machine translation of emergency communication preserves critical safety actionability. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Evaluates natural language emergency translation, not software engineering or coding.
SIREN: Situation-aware Index for Response and Emergency ActioNability in Crisis Translation This paper introduces an evaluation index to measure whether machine translation of emergency communication preserves critical safety actionability. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Deals with text translation in emergencies, with no multimodal modalities evaluated.
EHR-Bench: A Chinese Multi-Task Benchmark for Recruitment AI This paper introduces a Chinese multi-task evaluation benchmark covering six recruitment application areas for language models. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Presents an operational benchmark rather than focusing on safety, red teaming, or guardrails.
EHR-Bench: A Chinese Multi-Task Benchmark for Recruitment AI This paper introduces a Chinese multi-task evaluation benchmark covering six recruitment application areas for language models. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Presents a multi-task benchmark for recruitment tasks, not autonomous LLM agents.
EHR-Bench: A Chinese Multi-Task Benchmark for Recruitment AI This paper introduces a Chinese multi-task evaluation benchmark covering six recruitment application areas for language models. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Benchmark paper comparing models, not developing small or efficient architectures.
EHR-Bench: A Chinese Multi-Task Benchmark for Recruitment AI This paper introduces a Chinese multi-task evaluation benchmark covering six recruitment application areas for language models. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Includes a retrieval subtask, but its primary focus is a general recruitment evaluation benchmark.
EHR-Bench: A Chinese Multi-Task Benchmark for Recruitment AI This paper introduces a Chinese multi-task evaluation benchmark covering six recruitment application areas for language models. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses HR and recruitment workflows, not software development or coding assistance.
EHR-Bench: A Chinese Multi-Task Benchmark for Recruitment AI This paper introduces a Chinese multi-task evaluation benchmark covering six recruitment application areas for language models. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Evaluates text-based recruitment tasks without multimodal image, video, or audio data.
Attention-Graph Curvature for the Verification of Mathematical Reasoning This paper evaluates whether attention-graph curvature features can help verifiers detect errors in mathematical reasoning steps. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Adds computational overhead to forward passes rather than improving model efficiency.
Attention-Graph Curvature for the Verification of Mathematical Reasoning This paper evaluates whether attention-graph curvature features can help verifiers detect errors in mathematical reasoning steps. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Investigates mathematical reasoning steps rather than software coding or repair.
Attention-Graph Curvature for the Verification of Mathematical Reasoning This paper evaluates whether attention-graph curvature features can help verifiers detect errors in mathematical reasoning steps. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies internal attention curvature for step verification, not autonomous agentic systems.
Attention-Graph Curvature for the Verification of Mathematical Reasoning This paper evaluates whether attention-graph curvature features can help verifiers detect errors in mathematical reasoning steps. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on mathematical reasoning step verification, not retrieval-augmented generation.
Attention-Graph Curvature for the Verification of Mathematical Reasoning This paper evaluates whether attention-graph curvature features can help verifiers detect errors in mathematical reasoning steps. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Evaluates text-based mathematical reasoning verifiers, without multimodal inputs.
What Is the "Self" of a Large Language Model? This paper investigates internal model representations that separate assistant actions from externally provided text across conversational layers. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Deals strictly with text-based language model representations, not multimodal models.
What Is the "Self" of a Large Language Model? This paper investigates internal model representations that separate assistant actions from externally provided text across conversational layers. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Analyzes internal representations of conversational roles, not tool use or autonomous agents.
What Is the "Self" of a Large Language Model? This paper investigates internal model representations that separate assistant actions from externally provided text across conversational layers. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Investigates internal layer activations regarding self-representation, not document retrieval.
What Is the "Self" of a Large Language Model? This paper investigates internal model representations that separate assistant actions from externally provided text across conversational layers. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Presents mechanistic interpretability analysis rather than standard benchmarks or evaluation methodologies.
What Is the "Self" of a Large Language Model? This paper investigates internal model representations that separate assistant actions from externally provided text across conversational layers. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Examines geometric activation dynamics without focusing on compression or serving efficiency.
What Is the "Self" of a Large Language Model? This paper investigates internal model representations that separate assistant actions from externally provided text across conversational layers. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Studies internal activation representations of conversational roles, not code generation.
Learn from Move: the Next Step for GUI Agents This paper introduces an environment benchmark and exploration framework to help graphical user interface agents learn unfamiliar interaction dynamics. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on GUI interaction and exploration rather than retrieval-augmented generation.
Learn from Move: the Next Step for GUI Agents This paper introduces an environment benchmark and exploration framework to help graphical user interface agents learn unfamiliar interaction dynamics. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Fine-tunes a 9B model but does not focus on efficiency, compression, or quantization.
Learn from Move: the Next Step for GUI Agents This paper introduces an environment benchmark and exploration framework to help graphical user interface agents learn unfamiliar interaction dynamics. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Studies graphical user interface control, not software coding or program repair.
Learn from Move: the Next Step for GUI Agents This paper introduces an environment benchmark and exploration framework to help graphical user interface agents learn unfamiliar interaction dynamics. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 06:17	Not relevant Focuses on GUI agent exploration capability rather than safety guardrails or alignment.
Lost in Contrast: Instruction-Preserving Decoding for Reliable Visual Question Answering This paper introduces an instruction-preserving decoding strategy to prevent visual contrastive decoding from discarding valid abstention responses. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on contrastive decoding for VQA rather than external document retrieval.
Lost in Contrast: Instruction-Preserving Decoding for Reliable Visual Question Answering This paper introduces an instruction-preserving decoding strategy to prevent visual contrastive decoding from discarding valid abstention responses. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Proposes a decoding technique for visual question answering, not agentic planning or tools.
Lost in Contrast: Instruction-Preserving Decoding for Reliable Visual Question Answering This paper introduces an instruction-preserving decoding strategy to prevent visual contrastive decoding from discarding valid abstention responses. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Investigates decoding contrast strategies rather than model compression or inference acceleration.
Lost in Contrast: Instruction-Preserving Decoding for Reliable Visual Question Answering This paper introduces an instruction-preserving decoding strategy to prevent visual contrastive decoding from discarding valid abstention responses. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses visual question answering decoding, not programming or code tasks.
Lost in Contrast: Instruction-Preserving Decoding for Reliable Visual Question Answering This paper introduces an instruction-preserving decoding strategy to prevent visual contrastive decoding from discarding valid abstention responses. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Focuses on a decoding method rather than proposing a general LLM evaluation benchmark.
Lost in Contrast: Instruction-Preserving Decoding for Reliable Visual Question Answering This paper introduces an instruction-preserving decoding strategy to prevent visual contrastive decoding from discarding valid abstention responses. Preprint	Computer vision for industrial applications via arxiv · 15 Oct, 06:17	Not relevant Evaluates general VQA benchmarks, not industrial inspection or manufacturing defect detection.
The Multilingual Artificial Hivemind: Do All LLMs Think Alike in Every Language? This paper investigates output diversity and text homogenization across twenty-three open-weight language models and thirteen languages. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Analyzes multilingual open-ended text generation, not code generation or software engineering.
The Multilingual Artificial Hivemind: Do All LLMs Think Alike in Every Language? This paper investigates output diversity and text homogenization across twenty-three open-weight language models and thirteen languages. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Studies multilingual text diversity and collapse, not LLM agents or tool use.
The Multilingual Artificial Hivemind: Do All LLMs Think Alike in Every Language? This paper investigates output diversity and text homogenization across twenty-three open-weight language models and thirteen languages. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Does not study model compression, quantization, distillation, or efficient inference.
The Multilingual Artificial Hivemind: Do All LLMs Think Alike in Every Language? This paper investigates output diversity and text homogenization across twenty-three open-weight language models and thirteen languages. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Examines multilingual text generation without any visual or audio multimodal components.
The Multilingual Artificial Hivemind: Do All LLMs Think Alike in Every Language? This paper investigates output diversity and text homogenization across twenty-three open-weight language models and thirteen languages. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Focuses on generation diversity and homogenization across languages, not retrieval systems.
Is the Platform Part of the Measurement? A Protocol and Simulation Study for Cross-Platform Equivalence of Machine-Labelled Policy Sentiment This paper presents a psychometric framework to test whether automated sentiment classification remains equivalent across different social media platforms. Preprint	Small and efficient models via arxiv · 15 Oct, 06:17	Not relevant Does not examine model compression, distillation, or efficient inference.
Is the Platform Part of the Measurement? A Protocol and Simulation Study for Cross-Platform Equivalence of Machine-Labelled Policy Sentiment This paper presents a psychometric framework to test whether automated sentiment classification remains equivalent across different social media platforms. Preprint	AI agents via arxiv · 15 Oct, 06:17	Not relevant Focuses on social media sentiment analysis, not LLM agents.
Is the Platform Part of the Measurement? A Protocol and Simulation Study for Cross-Platform Equivalence of Machine-Labelled Policy Sentiment This paper presents a psychometric framework to test whether automated sentiment classification remains equivalent across different social media platforms. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Does not address retrieval-augmented generation pipelines.
Is the Platform Part of the Measurement? A Protocol and Simulation Study for Cross-Platform Equivalence of Machine-Labelled Policy Sentiment This paper presents a psychometric framework to test whether automated sentiment classification remains equivalent across different social media platforms. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Deals exclusively with social media text data.
Is the Platform Part of the Measurement? A Protocol and Simulation Study for Cross-Platform Equivalence of Machine-Labelled Policy Sentiment This paper presents a psychometric framework to test whether automated sentiment classification remains equivalent across different social media platforms. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Examines sentiment analysis on social media platforms, not code.
A Bethe-Ansatz Source of Early Token Correlations: A marginal-preserving construction for parallel proposals This paper constructs joint token probability laws from physics models to improve speculative verification in parallel language model decoding. Preprint	Multimodal models via arxiv · 15 Oct, 06:17	Not relevant Presents mathematical token generation theory, not multimodal models.
A Bethe-Ansatz Source of Early Token Correlations: A marginal-preserving construction for parallel proposals This paper constructs joint token probability laws from physics models to improve speculative verification in parallel language model decoding. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17	Not relevant Theoretical statistical physics paper, not an evaluation methodology.
A Bethe-Ansatz Source of Early Token Correlations: A marginal-preserving construction for parallel proposals This paper constructs joint token probability laws from physics models to improve speculative verification in parallel language model decoding. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17	Not relevant Does not address retrieval or external knowledge integration.

11,041–11,100 of 17,313← Previous Next →