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
Moral Combat: Exploiting Moral Foundations Theory to Jailbreak LLMs The paper introduces a multi-agent framework that reframes harmful requests using Moral Foundations Theory to test model jailbreak vulnerabilities. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Uses multi-agent framing merely as an attack tool rather than studying agent systems.
Moral Combat: Exploiting Moral Foundations Theory to Jailbreak LLMs The paper introduces a multi-agent framework that reframes harmful requests using Moral Foundations Theory to test model jailbreak vulnerabilities. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Studies text prompt reframing attacks, without multimodal data.
Moral Combat: Exploiting Moral Foundations Theory to Jailbreak LLMs The paper introduces a multi-agent framework that reframes harmful requests using Moral Foundations Theory to test model jailbreak vulnerabilities. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Addresses safety jailbreaks, not software engineering or code models.
Moral Combat: Exploiting Moral Foundations Theory to Jailbreak LLMs The paper introduces a multi-agent framework that reframes harmful requests using Moral Foundations Theory to test model jailbreak vulnerabilities. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Investigates jailbreak vulnerabilities rather than proposing evaluation standards.
Moral Combat: Exploiting Moral Foundations Theory to Jailbreak LLMs The paper introduces a multi-agent framework that reframes harmful requests using Moral Foundations Theory to test model jailbreak vulnerabilities. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Focuses on prompt jailbreaking, not knowledge retrieval or grounding.
Moral Combat: Exploiting Moral Foundations Theory to Jailbreak LLMs The paper introduces a multi-agent framework that reframes harmful requests using Moral Foundations Theory to test model jailbreak vulnerabilities. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on safety vulnerabilities, not model compression or efficiency.
PA1-OPD: One-Step Prefix-Augmented On-Policy Distillation for Language Model Pretraining The paper proposes a prefix-augmented distillation method providing one-step student branch feedback during language model pretraining. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Presents pretraining distillation methods, not multi-step agent systems.
PA1-OPD: One-Step Prefix-Augmented On-Policy Distillation for Language Model Pretraining The paper proposes a prefix-augmented distillation method providing one-step student branch feedback during language model pretraining. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Covers pretraining knowledge distillation, not external retrieval or RAG.
PA1-OPD: One-Step Prefix-Augmented On-Policy Distillation for Language Model Pretraining The paper proposes a prefix-augmented distillation method providing one-step student branch feedback during language model pretraining. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Presents a training method rather than an evaluation benchmark.
PA1-OPD: One-Step Prefix-Augmented On-Policy Distillation for Language Model Pretraining The paper proposes a prefix-augmented distillation method providing one-step student branch feedback during language model pretraining. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on language model training efficiency, not safety or guardrails.
PA1-OPD: One-Step Prefix-Augmented On-Policy Distillation for Language Model Pretraining The paper proposes a prefix-augmented distillation method providing one-step student branch feedback during language model pretraining. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses on text language model pretraining distillation.
PA1-OPD: One-Step Prefix-Augmented On-Policy Distillation for Language Model Pretraining The paper proposes a prefix-augmented distillation method providing one-step student branch feedback during language model pretraining. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Pretraining distillation technique, not specialized for code LLMs.
PN-Mix: Pair-Norm Guided Mixed-Precision Key Cache Quantization for Efficient Long-Output LLM Inference The paper proposes a pair-norm guided mixed-precision quantization method for key caches during long-output transformer inference. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Introduces a quantization algorithm, not an evaluation framework or dataset.
PN-Mix: Pair-Norm Guided Mixed-Precision Key Cache Quantization for Efficient Long-Output LLM Inference The paper proposes a pair-norm guided mixed-precision quantization method for key caches during long-output transformer inference. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Applies to decoder-only text language models without multimodal data.
PN-Mix: Pair-Norm Guided Mixed-Precision Key Cache Quantization for Efficient Long-Output LLM Inference The paper proposes a pair-norm guided mixed-precision quantization method for key caches during long-output transformer inference. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Covers inference hardware optimization, not safety vulnerabilities or robustness.
PN-Mix: Pair-Norm Guided Mixed-Precision Key Cache Quantization for Efficient Long-Output LLM Inference The paper proposes a pair-norm guided mixed-precision quantization method for key caches during long-output transformer inference. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Applies to long-output inference quantization, not document retrieval systems.
PN-Mix: Pair-Norm Guided Mixed-Precision Key Cache Quantization for Efficient Long-Output LLM Inference The paper proposes a pair-norm guided mixed-precision quantization method for key caches during long-output transformer inference. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Focuses on KV cache quantization mechanics, not tool-using agent architectures.
PN-Mix: Pair-Norm Guided Mixed-Precision Key Cache Quantization for Efficient Long-Output LLM Inference The paper proposes a pair-norm guided mixed-precision quantization method for key caches during long-output transformer inference. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant General LLM KV cache quantization method, not specific to software engineering.
Twenty Days Later: A Pre-Registered Test of History-Based Forecasts of Brand Mentions in Gemini Answers The paper reports a pre-registered test measuring the stability of history-based forecasts for model brand mentions over twenty days. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Investigates output stability rather than model efficiency or distillation.
Twenty Days Later: A Pre-Registered Test of History-Based Forecasts of Brand Mentions in Gemini Answers The paper reports a pre-registered test measuring the stability of history-based forecasts for model brand mentions over twenty days. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on commercial brand mention tracking, not safety risks or jailbreaks.
Twenty Days Later: A Pre-Registered Test of History-Based Forecasts of Brand Mentions in Gemini Answers The paper reports a pre-registered test measuring the stability of history-based forecasts for model brand mentions over twenty days. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Analyzes general model brand mentions, unrelated to software code tasks.
Twenty Days Later: A Pre-Registered Test of History-Based Forecasts of Brand Mentions in Gemini Answers The paper reports a pre-registered test measuring the stability of history-based forecasts for model brand mentions over twenty days. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Analyzes raw generative model outputs, not retrieval-augmented pipelines.
Twenty Days Later: A Pre-Registered Test of History-Based Forecasts of Brand Mentions in Gemini Answers The paper reports a pre-registered test measuring the stability of history-based forecasts for model brand mentions over twenty days. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Measures brand mention consistency, not agent planning or tool use.
Twenty Days Later: A Pre-Registered Test of History-Based Forecasts of Brand Mentions in Gemini Answers The paper reports a pre-registered test measuring the stability of history-based forecasts for model brand mentions over twenty days. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Analyzes text query answers only, without multimodal components.
Are We Really Measuring Forgetting? Pitfalls in Evaluating Machine Unlearning for Multimodal LLMs The paper identifies evaluation pitfalls in multimodal machine unlearning and introduces a benchmark grounded in real-world knowledge. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Focuses on evaluating unlearning protocols rather than model compression or acceleration.
Are We Really Measuring Forgetting? Pitfalls in Evaluating Machine Unlearning for Multimodal LLMs The paper identifies evaluation pitfalls in multimodal machine unlearning and introduces a benchmark grounded in real-world knowledge. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Examines multimodal unlearning benchmarks, not software or code assistants.
Are We Really Measuring Forgetting? Pitfalls in Evaluating Machine Unlearning for Multimodal LLMs The paper identifies evaluation pitfalls in multimodal machine unlearning and introduces a benchmark grounded in real-world knowledge. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Focuses on machine unlearning evaluation, not interactive multi-step agents.
Are We Really Measuring Forgetting? Pitfalls in Evaluating Machine Unlearning for Multimodal LLMs The paper identifies evaluation pitfalls in multimodal machine unlearning and introduces a benchmark grounded in real-world knowledge. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Studies parameter unlearning in MLLMs, not document retrieval systems.
J-tau: A Fully Localized Japanese Benchmark for Tool-Agent-User Interaction The paper introduces a localized Japanese benchmark for evaluating tool-using conversational agents in customer-support scenarios. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Introduces an evaluation benchmark rather than efficient model architectures or inference.
J-tau: A Fully Localized Japanese Benchmark for Tool-Agent-User Interaction The paper introduces a localized Japanese benchmark for evaluating tool-using conversational agents in customer-support scenarios. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Evaluates agent task execution, not safety vulnerabilities or adversarial attacks.
J-tau: A Fully Localized Japanese Benchmark for Tool-Agent-User Interaction The paper introduces a localized Japanese benchmark for evaluating tool-using conversational agents in customer-support scenarios. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Focuses on tool-agent interaction benchmarks, not retrieval systems or RAG pipelines.
J-tau: A Fully Localized Japanese Benchmark for Tool-Agent-User Interaction The paper introduces a localized Japanese benchmark for evaluating tool-using conversational agents in customer-support scenarios. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Evaluates text-based dialogue and tool interactions without multimodal capabilities.
J-tau: A Fully Localized Japanese Benchmark for Tool-Agent-User Interaction The paper introduces a localized Japanese benchmark for evaluating tool-using conversational agents in customer-support scenarios. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Focuses on customer support agents, not code generation or software repair.
Selective Reasoning for Typed Decision Models: When to Start, Read, and Stop The paper proposes an interface for typed decision models that selectively invokes and interrupts reasoning to balance accuracy and latency. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Examines internal reasoning scheduling, not external retrieval or knowledge grounding.
Selective Reasoning for Typed Decision Models: When to Start, Read, and Stop The paper proposes an interface for typed decision models that selectively invokes and interrupts reasoning to balance accuracy and latency. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on inference latency and compute trade-offs, not safety guardrails.
Selective Reasoning for Typed Decision Models: When to Start, Read, and Stop The paper proposes an interface for typed decision models that selectively invokes and interrupts reasoning to balance accuracy and latency. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes an interruptible reasoning mechanism rather than an evaluation benchmark.
Selective Reasoning for Typed Decision Models: When to Start, Read, and Stop The paper proposes an interface for typed decision models that selectively invokes and interrupts reasoning to balance accuracy and latency. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Targets text reasoning models without multimodal inputs.
Selective Reasoning for Typed Decision Models: When to Start, Read, and Stop The paper proposes an interface for typed decision models that selectively invokes and interrupts reasoning to balance accuracy and latency. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Focuses on logical reasoning interruption, not software engineering or code.
Selective Reasoning for Typed Decision Models: When to Start, Read, and Stop The paper proposes an interface for typed decision models that selectively invokes and interrupts reasoning to balance accuracy and latency. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Addresses internal test-time reasoning control, not tool-using agentic workflows.
Beyond Distribution Matching: Selective Margin Preservation for Supervised Fine-Tuning The paper introduces a soft margin preservation target for supervised fine-tuning to retain non-target token rankings. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Covers supervised fine-tuning loss targets, not autonomous planning or agents.
Beyond Distribution Matching: Selective Margin Preservation for Supervised Fine-Tuning The paper introduces a soft margin preservation target for supervised fine-tuning to retain non-target token rankings. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant General fine-tuning technique evaluated on math and code, not specialized code methods.
Beyond Distribution Matching: Selective Margin Preservation for Supervised Fine-Tuning The paper introduces a soft margin preservation target for supervised fine-tuning to retain non-target token rankings. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Addresses fine-tuning optimization, not safety, alignment, or hallucination prevention.
Beyond Distribution Matching: Selective Margin Preservation for Supervised Fine-Tuning The paper introduces a soft margin preservation target for supervised fine-tuning to retain non-target token rankings. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Proposes an SFT algorithm, not an evaluation framework or benchmark study.
Beyond Distribution Matching: Selective Margin Preservation for Supervised Fine-Tuning The paper introduces a soft margin preservation target for supervised fine-tuning to retain non-target token rankings. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses entirely on text-based language model supervised fine-tuning.
Beyond Distribution Matching: Selective Margin Preservation for Supervised Fine-Tuning The paper introduces a soft margin preservation target for supervised fine-tuning to retain non-target token rankings. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Focuses on supervised fine-tuning objectives, not retrieval systems.
Beyond Distribution Matching: Selective Margin Preservation for Supervised Fine-Tuning The paper introduces a soft margin preservation target for supervised fine-tuning to retain non-target token rankings. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Presents a fine-tuning loss rather than pruning, distillation, or quantization.
Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models The paper proposes a training-free interface for heterogeneous language models to communicate facts using shared symbolic word indices. Preprint	LLM evaluation and benchmarks via arxiv · 15 Oct, 15:06	Not relevant Introduces a communication protocol method rather than a general evaluation benchmark.
Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models The paper proposes a training-free interface for heterogeneous language models to communicate facts using shared symbolic word indices. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Focuses on cross-model hidden state symbol communication in text models.
Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models The paper proposes a training-free interface for heterogeneous language models to communicate facts using shared symbolic word indices. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Investigates multi-model hidden state sharing, not compression or inference optimization.
Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models The paper proposes a training-free interface for heterogeneous language models to communicate facts using shared symbolic word indices. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Studies inter-model state communication rather than document retrieval or RAG databases.
Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models The paper proposes a training-free interface for heterogeneous language models to communicate facts using shared symbolic word indices. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Explores neural communication between models, not code generation or analysis.
Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models The paper proposes a training-free interface for heterogeneous language models to communicate facts using shared symbolic word indices. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Focuses on collaborative multi-agent communication, not deployment safety or guardrails.
BANGLAPLAN: A Benchmark for Constraint-Based Planning and Decision-Making in Bangla The paper introduces a 1,000-problem benchmark for constraint-based planning and decision-making in the Bangla language. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant Evaluates natural language schedule/route planning, not program synthesis.
BANGLAPLAN: A Benchmark for Constraint-Based Planning and Decision-Making in Bangla The paper introduces a 1,000-problem benchmark for constraint-based planning and decision-making in the Bangla language. Preprint	Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 15:06	Not relevant Covers planning and constraint satisfaction benchmarks, not retrieval-augmented generation.
BANGLAPLAN: A Benchmark for Constraint-Based Planning and Decision-Making in Bangla The paper introduces a 1,000-problem benchmark for constraint-based planning and decision-making in the Bangla language. Preprint	AI safety and reliability in deployment via arxiv · 15 Oct, 15:06	Not relevant Measures task accuracy under constraints, not safety, jailbreaks, or robustness.
BANGLAPLAN: A Benchmark for Constraint-Based Planning and Decision-Making in Bangla The paper introduces a 1,000-problem benchmark for constraint-based planning and decision-making in the Bangla language. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Text-only planning benchmark in Bangla without vision or audio modalities.
BANGLAPLAN: A Benchmark for Constraint-Based Planning and Decision-Making in Bangla The paper introduces a 1,000-problem benchmark for constraint-based planning and decision-making in the Bangla language. Preprint	AI agents via arxiv · 15 Oct, 15:06	Not relevant Introduces a static problem benchmark rather than an interactive agent system.
BANGLAPLAN: A Benchmark for Constraint-Based Planning and Decision-Making in Bangla The paper introduces a 1,000-problem benchmark for constraint-based planning and decision-making in the Bangla language. Preprint	Small and efficient models via arxiv · 15 Oct, 15:06	Not relevant Presents an evaluation benchmark rather than model compression or efficient serving.
RouteWeaver: Weaving Mode Selection and Execution into Unified LLM Routing The paper introduces a unified routing framework that learns both execution mode selection and model assignment across diverse tasks. Preprint	Multimodal models via arxiv · 15 Oct, 15:06	Not relevant Addresses routing across language tasks without multimodal capabilities.
RouteWeaver: Weaving Mode Selection and Execution into Unified LLM Routing The paper introduces a unified routing framework that learns both execution mode selection and model assignment across diverse tasks. Preprint	LLMs for code via arxiv · 15 Oct, 15:06	Not relevant General LLM routing framework tested broadly, not specialized software engineering tools.

6,841–6,900 of 17,313← Previous Next →