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
P2TPO: Structure-Aligned Page-to-Token Credit Assignment for Full-Page Document Parsing Proposes a reinforcement learning credit-assignment method for document parsing models to align rewards with structured page elements. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Proposes a training credit assignment method, not an evaluation framework.
P2TPO: Structure-Aligned Page-to-Token Credit Assignment for Full-Page Document Parsing Proposes a reinforcement learning credit-assignment method for document parsing models to align rewards with structured page elements. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Does not focus on model compression, quantization, or inference speedup.
WEAVE: Weaving Real-World Code Evolution into Dependency-Structured Tasks for Benchmarking and Training Code Agents Introduces a benchmark that structures multi-pull-request code changes into dependency graphs to evaluate long-horizon software engineering agents. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Focuses on software engineering agents rather than retrieval-augmented generation systems.
WEAVE: Weaving Real-World Code Evolution into Dependency-Structured Tasks for Benchmarking and Training Code Agents Introduces a benchmark that structures multi-pull-request code changes into dependency graphs to evaluate long-horizon software engineering agents. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Examines code task progress rather than safety guardrails or vulnerabilities.
WEAVE: Weaving Real-World Code Evolution into Dependency-Structured Tasks for Benchmarking and Training Code Agents Introduces a benchmark that structures multi-pull-request code changes into dependency graphs to evaluate long-horizon software engineering agents. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Only deals with text and code repositories, not multimodal data.
WEAVE: Weaving Real-World Code Evolution into Dependency-Structured Tasks for Benchmarking and Training Code Agents Introduces a benchmark that structures multi-pull-request code changes into dependency graphs to evaluate long-horizon software engineering agents. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Does not address model compression, pruning, quantization, or efficient serving.
Effectiveness of Hidden-State Probes for Localizing Reasoning Failures Evaluates whether internal hidden-state probes can identify failure modes and pinpoint failing steps in multi-step reasoning traces. Peer-reviewed	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Tests 4B and 8B models but not model compression or efficiency.
Effectiveness of Hidden-State Probes for Localizing Reasoning Failures Evaluates whether internal hidden-state probes can identify failure modes and pinpoint failing steps in multi-step reasoning traces. Peer-reviewed	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Focuses solely on text-based language model reasoning without multimodal inputs.
Effectiveness of Hidden-State Probes for Localizing Reasoning Failures Evaluates whether internal hidden-state probes can identify failure modes and pinpoint failing steps in multi-step reasoning traces. Peer-reviewed	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Studies general mathematical and reasoning traces, not software engineering or code.
Effectiveness of Hidden-State Probes for Localizing Reasoning Failures Evaluates whether internal hidden-state probes can identify failure modes and pinpoint failing steps in multi-step reasoning traces. Peer-reviewed	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Does not involve external document retrieval, grounding, or RAG architectures.
Effectiveness of Hidden-State Probes for Localizing Reasoning Failures Evaluates whether internal hidden-state probes can identify failure modes and pinpoint failing steps in multi-step reasoning traces. Peer-reviewed	AI agents via arxiv · 17 Oct, 06:17	Not relevant Analyzes hidden-state probes in reasoning traces, not multi-step tool-using agents.
Name-Bias in Large Language Models Measures demographic bias related to age, gender, and race based on first names across local language models. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Tests local models for bias, not efficiency or compression.
Name-Bias in Large Language Models Measures demographic bias related to age, gender, and race based on first names across local language models. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Does not involve retrieval-augmented generation systems.
Name-Bias in Large Language Models Measures demographic bias related to age, gender, and race based on first names across local language models. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Studies demographic biases rather than agentic planning or tools.
Name-Bias in Large Language Models Measures demographic bias related to age, gender, and race based on first names across local language models. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Examines text prompt fill-in-the-blank responses only.
Name-Bias in Large Language Models Measures demographic bias related to age, gender, and race based on first names across local language models. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Does not cover code generation or software engineering.
Documenting Trofinetide Treatment Phenotypes in Rett Syndrome with a Dual-LLM Consensus Pipeline Applies a dual language model consensus pipeline to extract clinical phenotypes from electronic health records for patients with Rett syndrome. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Performs direct medical note extraction without external retrieval systems.
Documenting Trofinetide Treatment Phenotypes in Rett Syndrome with a Dual-LLM Consensus Pipeline Applies a dual language model consensus pipeline to extract clinical phenotypes from electronic health records for patients with Rett syndrome. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Does not study model compression, distillation, or serving speed.
Documenting Trofinetide Treatment Phenotypes in Rett Syndrome with a Dual-LLM Consensus Pipeline Applies a dual language model consensus pipeline to extract clinical phenotypes from electronic health records for patients with Rett syndrome. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Develops a clinical extraction pipeline, not an interactive agent.
Documenting Trofinetide Treatment Phenotypes in Rett Syndrome with a Dual-LLM Consensus Pipeline Applies a dual language model consensus pipeline to extract clinical phenotypes from electronic health records for patients with Rett syndrome. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Analyzes electronic clinical text notes, not multimodal data.
Documenting Trofinetide Treatment Phenotypes in Rett Syndrome with a Dual-LLM Consensus Pipeline Applies a dual language model consensus pipeline to extract clinical phenotypes from electronic health records for patients with Rett syndrome. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Focuses on a specific clinical application rather than general LLM evaluation.
Documenting Trofinetide Treatment Phenotypes in Rett Syndrome with a Dual-LLM Consensus Pipeline Applies a dual language model consensus pipeline to extract clinical phenotypes from electronic health records for patients with Rett syndrome. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Does not address code writing or software engineering.
Beyond Context Length: Characterising the Difficulty of Long-Context Reasoning Presents a benchmark that measures long-context reasoning difficulty using evidence scope, dispersion, and reasoning depth. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Evaluates reasoning difficulty rather than efficient inference or compression.
Beyond Context Length: Characterising the Difficulty of Long-Context Reasoning Presents a benchmark that measures long-context reasoning difficulty using evidence scope, dispersion, and reasoning depth. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Does not focus on software engineering or code generation.
Beyond Context Length: Characterising the Difficulty of Long-Context Reasoning Presents a benchmark that measures long-context reasoning difficulty using evidence scope, dispersion, and reasoning depth. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Focuses strictly on long-context textual reasoning.
Beyond Context Length: Characterising the Difficulty of Long-Context Reasoning Presents a benchmark that measures long-context reasoning difficulty using evidence scope, dispersion, and reasoning depth. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Evaluates long-context question-answering rather than agentic tool use.
Beyond Context Length: Characterising the Difficulty of Long-Context Reasoning Presents a benchmark that measures long-context reasoning difficulty using evidence scope, dispersion, and reasoning depth. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Studies context difficulty metrics, not safety or guardrails.
The Right Byte First: How UTF-16 Helps Multilingual Boundary Prediction Studies how using UTF-16 byte order improves boundary prediction and representation efficiency in multilingual byte-level language models. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Concerns text character encodings without multimodal inputs.
The Right Byte First: How UTF-16 Helps Multilingual Boundary Prediction Studies how using UTF-16 byte order improves boundary prediction and representation efficiency in multilingual byte-level language models. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Investigates byte encoding schemes, not agentic planning or execution.
The Right Byte First: How UTF-16 Helps Multilingual Boundary Prediction Studies how using UTF-16 byte order improves boundary prediction and representation efficiency in multilingual byte-level language models. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Does not address retrieval-augmented generation.
The Right Byte First: How UTF-16 Helps Multilingual Boundary Prediction Studies how using UTF-16 byte order improves boundary prediction and representation efficiency in multilingual byte-level language models. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Focuses on multilingual character encodings, not programming.
The Right Byte First: How UTF-16 Helps Multilingual Boundary Prediction Studies how using UTF-16 byte order improves boundary prediction and representation efficiency in multilingual byte-level language models. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Studies byte encoding schemes rather than model benchmarking.
The Right Byte First: How UTF-16 Helps Multilingual Boundary Prediction Studies how using UTF-16 byte order improves boundary prediction and representation efficiency in multilingual byte-level language models. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Does not touch on safety, jailbreaks, or deployment guardrails.
Patch where the text is new: a compression view of where byte-level models should spend compute Introduces a compression-based patching strategy for byte-level language models that allocates compute where upcoming bytes are unlikely to repeat. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Evaluates text and byte-level modeling without multimodal inputs.
Patch where the text is new: a compression view of where byte-level models should spend compute Introduces a compression-based patching strategy for byte-level language models that allocates compute where upcoming bytes are unlikely to repeat. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Focuses on patch allocation methods rather than evaluation benchmarking.
Patch where the text is new: a compression view of where byte-level models should spend compute Introduces a compression-based patching strategy for byte-level language models that allocates compute where upcoming bytes are unlikely to repeat. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Tests math and general bytes, not code assistant systems.
Patch where the text is new: a compression view of where byte-level models should spend compute Introduces a compression-based patching strategy for byte-level language models that allocates compute where upcoming bytes are unlikely to repeat. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Does not investigate safety, red-teaming, or model alignment.
Patch where the text is new: a compression view of where byte-level models should spend compute Introduces a compression-based patching strategy for byte-level language models that allocates compute where upcoming bytes are unlikely to repeat. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Does not involve external document retrieval or grounding.
Patch where the text is new: a compression view of where byte-level models should spend compute Introduces a compression-based patching strategy for byte-level language models that allocates compute where upcoming bytes are unlikely to repeat. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Studies byte-level tokenization mechanics, not autonomous agents.
Unfragmenting Nawatl NMT: Single-Variant Specialization for Masewalkopa Specializes a translation model to a specific indigenous Nawatl variety using a curated dataset and finite-state morphological analysis. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Focuses on indigenous language translation, not code generation.
Unfragmenting Nawatl NMT: Single-Variant Specialization for Masewalkopa Specializes a translation model to a specific indigenous Nawatl variety using a curated dataset and finite-state morphological analysis. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Studies machine translation specialization, not interactive AI agents.
Unfragmenting Nawatl NMT: Single-Variant Specialization for Masewalkopa Specializes a translation model to a specific indigenous Nawatl variety using a curated dataset and finite-state morphological analysis. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Only handles text translation, not multimodal data.
Unfragmenting Nawatl NMT: Single-Variant Specialization for Masewalkopa Specializes a translation model to a specific indigenous Nawatl variety using a curated dataset and finite-state morphological analysis. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Does not evaluate deployment safety risks or model robustness.
Unfragmenting Nawatl NMT: Single-Variant Specialization for Masewalkopa Specializes a translation model to a specific indigenous Nawatl variety using a curated dataset and finite-state morphological analysis. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Applies standard quantization to a translation model without core focus.
Unfragmenting Nawatl NMT: Single-Variant Specialization for Masewalkopa Specializes a translation model to a specific indigenous Nawatl variety using a curated dataset and finite-state morphological analysis. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Presents translation models rather than LLM evaluation benchmarks.
Unfragmenting Nawatl NMT: Single-Variant Specialization for Masewalkopa Specializes a translation model to a specific indigenous Nawatl variety using a curated dataset and finite-state morphological analysis. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Does not evaluate or use retrieval-augmented generation.
Intern-Decision: Efficient Multimodal Decision Making with Autoregressive Masked Language Modeling Develops small multimodal models that output structured choices using masked language modeling to reduce inference latency in agent tasks. Preprint	AI safety and reliability in deployment via arxiv · 17 Oct, 06:17	Not relevant Focuses on inference speed and calibration rather than safety guardrails.
Intern-Decision: Efficient Multimodal Decision Making with Autoregressive Masked Language Modeling Develops small multimodal models that output structured choices using masked language modeling to reduce inference latency in agent tasks. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Does not address code writing or software engineering tasks.
Intern-Decision: Efficient Multimodal Decision Making with Autoregressive Masked Language Modeling Develops small multimodal models that output structured choices using masked language modeling to reduce inference latency in agent tasks. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Does not study external retrieval or document augmentation.
Intern-Decision: Efficient Multimodal Decision Making with Autoregressive Masked Language Modeling Develops small multimodal models that output structured choices using masked language modeling to reduce inference latency in agent tasks. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Introduces a model architecture rather than an evaluation benchmark.
PremiseSQL: Evaluating Premise Critique and Abstention in Text-to-SQL Systems Introduces a benchmark to evaluate whether text-to-SQL systems can detect flawed query premises and abstain from generating invalid SQL. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Does not investigate model compression, distillation, or serving efficiency.
PremiseSQL: Evaluating Premise Critique and Abstention in Text-to-SQL Systems Introduces a benchmark to evaluate whether text-to-SQL systems can detect flawed query premises and abstain from generating invalid SQL. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Evaluates text-to-SQL generation rather than agentic tool workflows.
PremiseSQL: Evaluating Premise Critique and Abstention in Text-to-SQL Systems Introduces a benchmark to evaluate whether text-to-SQL systems can detect flawed query premises and abstain from generating invalid SQL. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Deals with text and SQL code, not multimodal media.
PremiseSQL: Evaluating Premise Critique and Abstention in Text-to-SQL Systems Introduces a benchmark to evaluate whether text-to-SQL systems can detect flawed query premises and abstain from generating invalid SQL. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Studies SQL generation from flawed natural language, not RAG.
CLUE: Uncovering Social Context for Pragmatic Understanding Develops a two-stage training approach combining evidence reasoning and reinforcement learning to improve pragmatic social reasoning in conversational models. Preprint	AI agents via arxiv · 17 Oct, 06:17	Not relevant Focuses on pragmatic conversation understanding rather than autonomous agents.
CLUE: Uncovering Social Context for Pragmatic Understanding Develops a two-stage training approach combining evidence reasoning and reinforcement learning to improve pragmatic social reasoning in conversational models. Preprint	Retrieval-augmented generation (RAG) via arxiv · 17 Oct, 06:17	Not relevant Does not involve external document retrieval or grounding architectures.
CLUE: Uncovering Social Context for Pragmatic Understanding Develops a two-stage training approach combining evidence reasoning and reinforcement learning to improve pragmatic social reasoning in conversational models. Preprint	LLMs for code via arxiv · 17 Oct, 06:17	Not relevant Not related to programming, software maintenance, or code generation.
CLUE: Uncovering Social Context for Pragmatic Understanding Develops a two-stage training approach combining evidence reasoning and reinforcement learning to improve pragmatic social reasoning in conversational models. Preprint	LLM evaluation and benchmarks via arxiv · 17 Oct, 06:17	Not relevant Proposes a training pipeline rather than a standardized evaluation framework.
CLUE: Uncovering Social Context for Pragmatic Understanding Develops a two-stage training approach combining evidence reasoning and reinforcement learning to improve pragmatic social reasoning in conversational models. Preprint	Small and efficient models via arxiv · 17 Oct, 06:17	Not relevant Reduces reasoning tokens via prompting/RL, not model compression.
CLUE: Uncovering Social Context for Pragmatic Understanding Develops a two-stage training approach combining evidence reasoning and reinforcement learning to improve pragmatic social reasoning in conversational models. Preprint	Multimodal models via arxiv · 17 Oct, 06:17	Not relevant Deals exclusively with text dialogue understanding.

4,741–4,800 of 17,313← Previous Next →