Every paper checked — nothing hidden
All papers
Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.
All17,313In brief24More (on-topic)3,333Not relevant13,754Not assessed87Older352Could not confirm93Blogs with no paper266All blogs284
| Paper | Topic | Relevance | Feedback |
|---|---|---|---|
| Gluing Local Contexts into Global Meaning: A Sheaf-Theoretic Decomposition of Transformer Representations This paper proposes a mathematical framework using cellular sheaves to isolate phrasing-invariant content representations in transformer language models. Preprint | Multimodal models via arxiv · 15 Oct, 06:17 | Not relevant Examines text transformer representations without multimodal components. | |
| Gluing Local Contexts into Global Meaning: A Sheaf-Theoretic Decomposition of Transformer Representations This paper proposes a mathematical framework using cellular sheaves to isolate phrasing-invariant content representations in transformer language models. Preprint | Small and efficient models via arxiv · 15 Oct, 06:17 | Not relevant Examines internal representation geometry, not model compression or efficiency. | |
| Gluing Local Contexts into Global Meaning: A Sheaf-Theoretic Decomposition of Transformer Representations This paper proposes a mathematical framework using cellular sheaves to isolate phrasing-invariant content representations in transformer language models. Preprint | LLMs for code via arxiv · 15 Oct, 06:17 | Not relevant Studies internal representations of language models, not software development. | |
| Gluing Local Contexts into Global Meaning: A Sheaf-Theoretic Decomposition of Transformer Representations This paper proposes a mathematical framework using cellular sheaves to isolate phrasing-invariant content representations in transformer language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an interpretability framework rather than benchmark evaluation. | |
| Gluing Local Contexts into Global Meaning: A Sheaf-Theoretic Decomposition of Transformer Representations This paper proposes a mathematical framework using cellular sheaves to isolate phrasing-invariant content representations in transformer language models. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17 | Not relevant Does not deal with retrieval-augmented generation architectures. | |
| Gluing Local Contexts into Global Meaning: A Sheaf-Theoretic Decomposition of Transformer Representations This paper proposes a mathematical framework using cellular sheaves to isolate phrasing-invariant content representations in transformer language models. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | Not relevant Theoretical interpretability study not directly focused on safety guardrails or robustness. | |
| ToolHACE: A Dataset for Detecting Hallucinations in Tool-Based LLM Responses This paper introduces a benchmark dataset to evaluate hallucination detection in language model responses after external tool execution. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17 | Not relevant Focuses on tool-calling outputs in dialogue rather than document retrieval or RAG. | |
| ToolHACE: A Dataset for Detecting Hallucinations in Tool-Based LLM Responses This paper introduces a benchmark dataset to evaluate hallucination detection in language model responses after external tool execution. Preprint | LLMs for code via arxiv · 15 Oct, 06:17 | Not relevant Focuses on general tool-augmented dialogue, not code generation or repair. | |
| ToolHACE: A Dataset for Detecting Hallucinations in Tool-Based LLM Responses This paper introduces a benchmark dataset to evaluate hallucination detection in language model responses after external tool execution. Preprint | Small and efficient models via arxiv · 15 Oct, 06:17 | Not relevant Does not focus on small models, quantization, distillation, or serving efficiency. | |
| ToolHACE: A Dataset for Detecting Hallucinations in Tool-Based LLM Responses This paper introduces a benchmark dataset to evaluate hallucination detection in language model responses after external tool execution. Preprint | Multimodal models via arxiv · 15 Oct, 06:17 | Not relevant Deals with text and tool responses, not multimodal vision/audio tasks. | |
| DuoDiscover: Scaling Autonomous Discovery Beyond the Evaluation Bottleneck The paper proposes an agent framework that discovers lightweight proxy evaluators to bypass costly benchmark evaluations during autonomous engineering discovery. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | Not relevant Addresses optimization efficiency rather than safety risks, jailbreaks, or deployment guardrails. | |
| DuoDiscover: Scaling Autonomous Discovery Beyond the Evaluation Bottleneck The paper proposes an agent framework that discovers lightweight proxy evaluators to bypass costly benchmark evaluations during autonomous engineering discovery. Preprint | Small and efficient models via arxiv · 15 Oct, 06:17 | Not relevant Addresses search and evaluation scaling rather than model compression or efficient serving. | |
| DuoDiscover: Scaling Autonomous Discovery Beyond the Evaluation Bottleneck The paper proposes an agent framework that discovers lightweight proxy evaluators to bypass costly benchmark evaluations during autonomous engineering discovery. Preprint | Multimodal models via arxiv · 15 Oct, 06:17 | Not relevant Focuses on software engineering code agents, not multimodal understanding. | |
| DuoDiscover: Scaling Autonomous Discovery Beyond the Evaluation Bottleneck The paper proposes an agent framework that discovers lightweight proxy evaluators to bypass costly benchmark evaluations during autonomous engineering discovery. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17 | Not relevant Does not address retrieval-augmented generation or external document indexing. | |
| Soft Teacher Forcing The paper proposes combining teacher-forced embeddings with soft tokens to improve supervised fine-tuning on reasoning datasets without changing inference. Preprint | AI agents via arxiv · 15 Oct, 06:17 | Not relevant Focuses on fine-tuning training loss, not agent workflows or multi-step execution. | |
| Soft Teacher Forcing The paper proposes combining teacher-forced embeddings with soft tokens to improve supervised fine-tuning on reasoning datasets without changing inference. Preprint | Small and efficient models via arxiv · 15 Oct, 06:17 | Not relevant Studies supervised fine-tuning training efficiency, not model compression or serving. | |
| Soft Teacher Forcing The paper proposes combining teacher-forced embeddings with soft tokens to improve supervised fine-tuning on reasoning datasets without changing inference. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Proposes a training method, not a benchmark or evaluation methodology. | |
| Soft Teacher Forcing The paper proposes combining teacher-forced embeddings with soft tokens to improve supervised fine-tuning on reasoning datasets without changing inference. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17 | Not relevant Does not involve retrieval, external knowledge sources, or grounding. | |
| Soft Teacher Forcing The paper proposes combining teacher-forced embeddings with soft tokens to improve supervised fine-tuning on reasoning datasets without changing inference. Preprint | Multimodal models via arxiv · 15 Oct, 06:17 | Not relevant Deals entirely with text-based language models, not multimodal inputs. | |
| Soft Teacher Forcing The paper proposes combining teacher-forced embeddings with soft tokens to improve supervised fine-tuning on reasoning datasets without changing inference. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | Not relevant Addresses training techniques for reasoning, not safety, guardrails, or alignment. | |
| Soft Teacher Forcing The paper proposes combining teacher-forced embeddings with soft tokens to improve supervised fine-tuning on reasoning datasets without changing inference. Preprint | LLMs for code via arxiv · 15 Oct, 06:17 | Not relevant Evaluated on math datasets, not code generation or software engineering. | |
| Welcome to the Age of Agent Ads: Using On-Policy Attacks to Steer Agentic Decisions The paper investigates on-policy indirect manipulation attacks that steer autonomous agent purchasing and ranking decisions via web retrieval. Preprint | Multimodal models via arxiv · 15 Oct, 06:17 | Not relevant Deals with text-based agent web browsing and ranking, not multimodal models. | |
| Welcome to the Age of Agent Ads: Using On-Policy Attacks to Steer Agentic Decisions The paper investigates on-policy indirect manipulation attacks that steer autonomous agent purchasing and ranking decisions via web retrieval. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on adversarial steering attacks rather than LLM benchmarking methods. | |
| Welcome to the Age of Agent Ads: Using On-Policy Attacks to Steer Agentic Decisions The paper investigates on-policy indirect manipulation attacks that steer autonomous agent purchasing and ranking decisions via web retrieval. Preprint | Small and efficient models via arxiv · 15 Oct, 06:17 | Not relevant Does not address model compression or inference serving efficiency. | |
| Welcome to the Age of Agent Ads: Using On-Policy Attacks to Steer Agentic Decisions The paper investigates on-policy indirect manipulation attacks that steer autonomous agent purchasing and ranking decisions via web retrieval. Preprint | LLMs for code via arxiv · 15 Oct, 06:17 | Not relevant Not related to software engineering or code generation. | |
| MLAfication: Efficiently Enabling High-Fidelity MLA for Modern LLMs The paper introduces an attention conversion method enabling multi-head latent attention on modern architectures with reduced memory footprints. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Presents an architecture optimization method rather than an evaluation benchmark. | |
| MLAfication: Efficiently Enabling High-Fidelity MLA for Modern LLMs The paper introduces an attention conversion method enabling multi-head latent attention on modern architectures with reduced memory footprints. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17 | Not relevant Does not address retrieval-augmented generation architectures. | |
| MLAfication: Efficiently Enabling High-Fidelity MLA for Modern LLMs The paper introduces an attention conversion method enabling multi-head latent attention on modern architectures with reduced memory footprints. Preprint | LLMs for code via arxiv · 15 Oct, 06:17 | Not relevant Addresses general LLM attention architecture rather than code generation. | |
| MLAfication: Efficiently Enabling High-Fidelity MLA for Modern LLMs The paper introduces an attention conversion method enabling multi-head latent attention on modern architectures with reduced memory footprints. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | Not relevant Addresses memory and throughput efficiency, not AI safety or alignment. | |
| MLAfication: Efficiently Enabling High-Fidelity MLA for Modern LLMs The paper introduces an attention conversion method enabling multi-head latent attention on modern architectures with reduced memory footprints. Preprint | AI agents via arxiv · 15 Oct, 06:17 | Not relevant Focuses on attention compression rather than multi-step agent execution. | |
| MLAfication: Efficiently Enabling High-Fidelity MLA for Modern LLMs The paper introduces an attention conversion method enabling multi-head latent attention on modern architectures with reduced memory footprints. Preprint | Multimodal models via arxiv · 15 Oct, 06:17 | Not relevant Focuses on core transformer attention efficiency, not multimodal integration. | |
| Online LLM Personalization from Pairwise Preferences Across Users The paper formulates online language model persona personalization as a dueling bandit problem using sequential pairwise user preferences. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | Not relevant Focuses on style personalization rather than safety, guardrails, or alignment risks. | |
| Online LLM Personalization from Pairwise Preferences Across Users The paper formulates online language model persona personalization as a dueling bandit problem using sequential pairwise user preferences. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17 | Not relevant Focuses on style personalization rather than knowledge retrieval or RAG. | |
| Online LLM Personalization from Pairwise Preferences Across Users The paper formulates online language model persona personalization as a dueling bandit problem using sequential pairwise user preferences. Preprint | Small and efficient models via arxiv · 15 Oct, 06:17 | Not relevant Does not focus on model compression, pruning, or serving efficiency. | |
| Online LLM Personalization from Pairwise Preferences Across Users The paper formulates online language model persona personalization as a dueling bandit problem using sequential pairwise user preferences. Preprint | LLMs for code via arxiv · 15 Oct, 06:17 | Not relevant Not related to code generation or software engineering. | |
| Online LLM Personalization from Pairwise Preferences Across Users The paper formulates online language model persona personalization as a dueling bandit problem using sequential pairwise user preferences. Preprint | AI agents via arxiv · 15 Oct, 06:17 | Not relevant Investigates prompt persona selection via dueling bandits, not agentic planning or tools. | |
| Online LLM Personalization from Pairwise Preferences Across Users The paper formulates online language model persona personalization as a dueling bandit problem using sequential pairwise user preferences. Preprint | Multimodal models via arxiv · 15 Oct, 06:17 | Not relevant Deals with text tutoring prompts, not multimodal applications. | |
| Nufo: A Benchmark for Ewe Automatic Speech Recognition The paper introduces a 174-hour speech benchmark and evaluates baseline models on Ewe automatic speech recognition. Preprint | Multimodal models via arxiv · 15 Oct, 06:17 | Not relevant Speech recognition dataset, outside the scope of multimodal vision-language models. | |
| Nufo: A Benchmark for Ewe Automatic Speech Recognition The paper introduces a 174-hour speech benchmark and evaluates baseline models on Ewe automatic speech recognition. Preprint | Small and efficient models via arxiv · 15 Oct, 06:17 | Not relevant Does not address model compression or efficient inference techniques. | |
| Nufo: A Benchmark for Ewe Automatic Speech Recognition The paper introduces a 174-hour speech benchmark and evaluates baseline models on Ewe automatic speech recognition. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | Not relevant Does not address AI safety, jailbreaks, or deployment guardrails. | |
| Nufo: A Benchmark for Ewe Automatic Speech Recognition The paper introduces a 174-hour speech benchmark and evaluates baseline models on Ewe automatic speech recognition. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Focuses on speech recognition benchmarking rather than language model evaluation. | |
| Nufo: A Benchmark for Ewe Automatic Speech Recognition The paper introduces a 174-hour speech benchmark and evaluates baseline models on Ewe automatic speech recognition. Preprint | LLMs for code via arxiv · 15 Oct, 06:17 | Not relevant Unrelated to software engineering or programming tasks. | |
| Nufo: A Benchmark for Ewe Automatic Speech Recognition The paper introduces a 174-hour speech benchmark and evaluates baseline models on Ewe automatic speech recognition. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17 | Not relevant Does not involve retrieval-augmented generation. | |
| Nufo: A Benchmark for Ewe Automatic Speech Recognition The paper introduces a 174-hour speech benchmark and evaluates baseline models on Ewe automatic speech recognition. Preprint | AI agents via arxiv · 15 Oct, 06:17 | Not relevant Covers low-resource speech recognition, not autonomous agent workflows. | |
| From Sampled Evidence to Radiology Reports: Objectives, Constraints, and Realization The paper evaluates how objective functions and candidate constraints affect decision conversion when generating radiology reports from sampled evidence. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17 | Not relevant Focuses on evidence aggregation and realization, not external document retrieval. | |
| From Sampled Evidence to Radiology Reports: Objectives, Constraints, and Realization The paper evaluates how objective functions and candidate constraints affect decision conversion when generating radiology reports from sampled evidence. Preprint | Small and efficient models via arxiv · 15 Oct, 06:17 | Not relevant Does not investigate model compression, quantization, or inference serving. | |
| From Sampled Evidence to Radiology Reports: Objectives, Constraints, and Realization The paper evaluates how objective functions and candidate constraints affect decision conversion when generating radiology reports from sampled evidence. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Studies report realization trade-offs rather than general LLM benchmarking. | |
| From Sampled Evidence to Radiology Reports: Objectives, Constraints, and Realization The paper evaluates how objective functions and candidate constraints affect decision conversion when generating radiology reports from sampled evidence. Preprint | AI agents via arxiv · 15 Oct, 06:17 | Not relevant Studies clinical report generation pipelines rather than tool-using autonomous agents. | |
| From Sampled Evidence to Radiology Reports: Objectives, Constraints, and Realization The paper evaluates how objective functions and candidate constraints affect decision conversion when generating radiology reports from sampled evidence. Preprint | LLMs for code via arxiv · 15 Oct, 06:17 | Not relevant Not focused on code generation or software engineering. | |
| The Curse of Knowledge in Role-Playing Language Models The paper tests whether role-playing language models leak private or unobserved character information across simulated interactive scenes. Preprint | LLMs for code via arxiv · 15 Oct, 06:17 | Not relevant Not related to software engineering or code generation. | |
| The Curse of Knowledge in Role-Playing Language Models The paper tests whether role-playing language models leak private or unobserved character information across simulated interactive scenes. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:17 | Not relevant Investigates information leakage in role-play, not RAG or document retrieval. | |
| The Curse of Knowledge in Role-Playing Language Models The paper tests whether role-playing language models leak private or unobserved character information across simulated interactive scenes. Preprint | Small and efficient models via arxiv · 15 Oct, 06:17 | Not relevant Does not address model compression or efficient serving. | |
| The Curse of Knowledge in Role-Playing Language Models The paper tests whether role-playing language models leak private or unobserved character information across simulated interactive scenes. Preprint | Multimodal models via arxiv · 15 Oct, 06:17 | Not relevant Focuses on text-based role-playing prompts, not multimodal models. | |
| UMCR: Universal Medical Concept Recognition through Guideline Integration and Agentic Linking The paper introduces a framework combining guidelines and agentic linking to ground medical concepts across diverse text types and ontologies. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | Not relevant Aimed at ontology linking accuracy rather than safety guardrails or jailbreaks. | |
| UMCR: Universal Medical Concept Recognition through Guideline Integration and Agentic Linking The paper introduces a framework combining guidelines and agentic linking to ground medical concepts across diverse text types and ontologies. Preprint | LLMs for code via arxiv · 15 Oct, 06:17 | Not relevant Focuses on clinical medical concept recognition rather than code. | |
| UMCR: Universal Medical Concept Recognition through Guideline Integration and Agentic Linking The paper introduces a framework combining guidelines and agentic linking to ground medical concepts across diverse text types and ontologies. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | Not relevant Applies existing MedPath benchmark rather than proposing LLM evaluation methods. | |
| UMCR: Universal Medical Concept Recognition through Guideline Integration and Agentic Linking The paper introduces a framework combining guidelines and agentic linking to ground medical concepts across diverse text types and ontologies. Preprint | Small and efficient models via arxiv · 15 Oct, 06:17 | Not relevant Does not focus on model compression, quantization, or serving efficiency. | |
| UMCR: Universal Medical Concept Recognition through Guideline Integration and Agentic Linking The paper introduces a framework combining guidelines and agentic linking to ground medical concepts across diverse text types and ontologies. Preprint | Multimodal models via arxiv · 15 Oct, 06:17 | Not relevant Processes clinical text and ontologies, not multimodal vision or audio. | |
| PIG: Parallel Inconsistency Guidance for Accelerating Diffusion Language Models The paper introduces a lightweight guidance model to estimate parallel decoding inconsistency and accelerate diffusion language models. Preprint | LLMs for code via arxiv · 15 Oct, 06:17 | Not relevant Evaluates general decoding speedups, not specialized code assistants. | |
| PIG: Parallel Inconsistency Guidance for Accelerating Diffusion Language Models The paper introduces a lightweight guidance model to estimate parallel decoding inconsistency and accelerate diffusion language models. Preprint | Multimodal models via arxiv · 15 Oct, 06:17 | Not relevant Applies to diffusion language models on text, not multimodal models. |
7,681–7,740 of 13,754← PreviousNext →