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
TokenRouter: Efficient Serving System for Token-Level LLM Routing Designs an inference serving system optimized for dynamic, token-level routing across multiple language models. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Addresses system throughput and serving efficiency rather than AI safety or guardrails.
NativeScope: Relation-Localized Retrieval over Native Topology with a Correct Anchor Proposes a scope-then-rank retrieval method using document topology and relational structure rather than pure semantic similarity. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on retrieval precision rather than AI safety or reliability.
NativeScope: Relation-Localized Retrieval over Native Topology with a Correct Anchor Proposes a scope-then-rank retrieval method using document topology and relational structure rather than pure semantic similarity. Preprint	LLMs for code via arxiv · 12 Oct, 16:28	Not relevant Addresses document retrieval rather than code generation or software engineering.
NativeScope: Relation-Localized Retrieval over Native Topology with a Correct Anchor Proposes a scope-then-rank retrieval method using document topology and relational structure rather than pure semantic similarity. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses on document retrieval ranking rather than autonomous agent systems.
NativeScope: Relation-Localized Retrieval over Native Topology with a Correct Anchor Proposes a scope-then-rank retrieval method using document topology and relational structure rather than pure semantic similarity. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Applies to text retrieval over document structure without multimodal data.
NativeScope: Relation-Localized Retrieval over Native Topology with a Correct Anchor Proposes a scope-then-rank retrieval method using document topology and relational structure rather than pure semantic similarity. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Focuses on retrieval algorithms rather than model compression or efficient inference.
NativeScope: Relation-Localized Retrieval over Native Topology with a Correct Anchor Proposes a scope-then-rank retrieval method using document topology and relational structure rather than pure semantic similarity. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Evaluates a specific retrieval algorithm rather than establishing a broad LLM benchmark.
EgoVoice: Proactive Spoken Assistance from Egocentric Multimodal Streams Develops a framework for egocentric multimodal assistants to decide both when to speak and what guidance to provide. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Addresses proactive assistance timing rather than deployment safety, jailbreaks, or guardrails.
EgoVoice: Proactive Spoken Assistance from Egocentric Multimodal Streams Develops a framework for egocentric multimodal assistants to decide both when to speak and what guidance to provide. Preprint	Computer vision for industrial applications via arxiv · 12 Oct, 16:28	Not relevant Designed for wearable consumer AR assistance rather than industrial manufacturing or inspection.
EgoVoice: Proactive Spoken Assistance from Egocentric Multimodal Streams Develops a framework for egocentric multimodal assistants to decide both when to speak and what guidance to provide. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Proposes a model framework rather than general LLM evaluation benchmarks.
EgoVoice: Proactive Spoken Assistance from Egocentric Multimodal Streams Develops a framework for egocentric multimodal assistants to decide both when to speak and what guidance to provide. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Focuses on multimodal interactive assistance rather than model compression or efficient serving.
EgoVoice: Proactive Spoken Assistance from Egocentric Multimodal Streams Develops a framework for egocentric multimodal assistants to decide both when to speak and what guidance to provide. Preprint	LLMs for code via arxiv · 12 Oct, 16:28	Not relevant Does not address programming, code generation, or software engineering.
EgoVoice: Proactive Spoken Assistance from Egocentric Multimodal Streams Develops a framework for egocentric multimodal assistants to decide both when to speak and what guidance to provide. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Focuses on egocentric streaming multimodal assistance rather than retrieval-augmented generation.
HarnessSQL: Harness-Native Training for SQL Agents in Realistic Database Environments Introduces a post-training framework that trains text-to-SQL agents directly inside interactive database execution environments. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Presents a training framework rather than proposing a general LLM benchmark.
HarnessSQL: Harness-Native Training for SQL Agents in Realistic Database Environments Introduces a post-training framework that trains text-to-SQL agents directly inside interactive database execution environments. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on execution accuracy in SQL tasks rather than safety, guardrails, or alignment.
HarnessSQL: Harness-Native Training for SQL Agents in Realistic Database Environments Introduces a post-training framework that trains text-to-SQL agents directly inside interactive database execution environments. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Deals exclusively with text and SQL code without multimodal processing.
HarnessSQL: Harness-Native Training for SQL Agents in Realistic Database Environments Introduces a post-training framework that trains text-to-SQL agents directly inside interactive database execution environments. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Addresses agent training for SQL tasks rather than model compression or serving efficiency.
HarnessSQL: Harness-Native Training for SQL Agents in Realistic Database Environments Introduces a post-training framework that trains text-to-SQL agents directly inside interactive database execution environments. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Deals with interactive database querying rather than document retrieval-augmented generation.
SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference Identifies distribution shift during autoregressive decoding and proposes a decoding-aware pruning method to accelerate model inference. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Proposes an inference pruning technique rather than an evaluation benchmark.
SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference Identifies distribution shift during autoregressive decoding and proposes a decoding-aware pruning method to accelerate model inference. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on inference latency rather than AI safety, jailbreaks, or alignment.
SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference Identifies distribution shift during autoregressive decoding and proposes a decoding-aware pruning method to accelerate model inference. Preprint	LLMs for code via arxiv · 12 Oct, 16:28	Not relevant Applies to LLM decoding efficiency broadly, not software engineering or coding tasks.
SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference Identifies distribution shift during autoregressive decoding and proposes a decoding-aware pruning method to accelerate model inference. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses on model pruning for decoding speedup rather than agent workflows.
SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference Identifies distribution shift during autoregressive decoding and proposes a decoding-aware pruning method to accelerate model inference. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Addresses weight pruning during decoding rather than retrieval-augmented generation.
SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference Identifies distribution shift during autoregressive decoding and proposes a decoding-aware pruning method to accelerate model inference. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Optimizes language model inference without multimodal inputs.
VFold: Symmetry-Aware Cross-Layer Value Cache Compression Presents a symmetry-aware value cache merging method to compress key-value memory during long-context decoding without architectural changes. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses on KV cache memory optimization rather than LLM agent architectures.
VFold: Symmetry-Aware Cross-Layer Value Cache Compression Presents a symmetry-aware value cache merging method to compress key-value memory during long-context decoding without architectural changes. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Addresses hardware efficiency and memory usage rather than AI safety or reliability.
VFold: Symmetry-Aware Cross-Layer Value Cache Compression Presents a symmetry-aware value cache merging method to compress key-value memory during long-context decoding without architectural changes. Preprint	LLMs for code via arxiv · 12 Oct, 16:28	Not relevant Focuses on general LLM memory efficiency rather than code generation or program repair.
VFold: Symmetry-Aware Cross-Layer Value Cache Compression Presents a symmetry-aware value cache merging method to compress key-value memory during long-context decoding without architectural changes. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Presents a systems optimization technique rather than an evaluation benchmark.
VFold: Symmetry-Aware Cross-Layer Value Cache Compression Presents a symmetry-aware value cache merging method to compress key-value memory during long-context decoding without architectural changes. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Focuses on transformer memory architecture without multimodal data.
VFold: Symmetry-Aware Cross-Layer Value Cache Compression Presents a symmetry-aware value cache merging method to compress key-value memory during long-context decoding without architectural changes. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Optimizes transformer KV cache compression rather than retrieval-augmented generation.
Which Skill to Distill? SGUID: Selecting a Compact Skill Bank for Model-Skill Co-Evolution Proposes a method to select a compact subset of procedural skills that provide effective training signals during model distillation. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Focuses on knowledge distillation of procedural guidance rather than autonomous agent systems.
Which Skill to Distill? SGUID: Selecting a Compact Skill Bank for Model-Skill Co-Evolution Proposes a method to select a compact subset of procedural skills that provide effective training signals during model distillation. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Focuses on language model distillation without multimodal modalities.
Which Skill to Distill? SGUID: Selecting a Compact Skill Bank for Model-Skill Co-Evolution Proposes a method to select a compact subset of procedural skills that provide effective training signals during model distillation. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Studies selecting skills for model distillation rather than run-time retrieval-augmented generation.
Which Skill to Distill? SGUID: Selecting a Compact Skill Bank for Model-Skill Co-Evolution Proposes a method to select a compact subset of procedural skills that provide effective training signals during model distillation. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Introduces a distillation method rather than a general LLM evaluation benchmark.
Which Skill to Distill? SGUID: Selecting a Compact Skill Bank for Model-Skill Co-Evolution Proposes a method to select a compact subset of procedural skills that provide effective training signals during model distillation. Preprint	LLMs for code via arxiv · 12 Oct, 16:28	Not relevant Investigates general skill distillation rather than software development or coding models.
Which Skill to Distill? SGUID: Selecting a Compact Skill Bank for Model-Skill Co-Evolution Proposes a method to select a compact subset of procedural skills that provide effective training signals during model distillation. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on distillation performance rather than AI safety, jailbreaks, or deployment guardrails.
Latent Core Tokenizer: Compress, but Meaningfully Introduces a multilingual tokenizer that uses structural discovery and linguistic constraints to improve downstream model representations. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Focuses strictly on multilingual text tokenization without multimodal components.
Latent Core Tokenizer: Compress, but Meaningfully Introduces a multilingual tokenizer that uses structural discovery and linguistic constraints to improve downstream model representations. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Investigates tokenizer construction methods rather than multi-step LLM agent systems.
Latent Core Tokenizer: Compress, but Meaningfully Introduces a multilingual tokenizer that uses structural discovery and linguistic constraints to improve downstream model representations. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Focuses on subword tokenization algorithms rather than retrieval or grounding mechanisms.
Latent Core Tokenizer: Compress, but Meaningfully Introduces a multilingual tokenizer that uses structural discovery and linguistic constraints to improve downstream model representations. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Introduces a tokenization method rather than an LLM evaluation methodology or benchmark.
Latent Core Tokenizer: Compress, but Meaningfully Introduces a multilingual tokenizer that uses structural discovery and linguistic constraints to improve downstream model representations. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Studies linguistic compression rather than AI safety, guardrails, or deployment reliability.
Latent Core Tokenizer: Compress, but Meaningfully Introduces a multilingual tokenizer that uses structural discovery and linguistic constraints to improve downstream model representations. Preprint	LLMs for code via arxiv · 12 Oct, 16:28	Not relevant Studies multilingual vocabulary tokenization without addressing software engineering or code LLMs.
Long Text to Predictive Features: LLM-Guided Blockwise Feature Engineering via Executable Program Search Presents a framework that uses language models offline to generate executable feature-engineering code for tabular risk models. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Presents a feature engineering method rather than an LLM evaluation framework or benchmark.
Long Text to Predictive Features: LLM-Guided Blockwise Feature Engineering via Executable Program Search Presents a framework that uses language models offline to generate executable feature-engineering code for tabular risk models. Preprint	Multimodal models via arxiv · 12 Oct, 16:28	Not relevant Focuses on text processing and program generation without multimodal inputs.
Long Text to Predictive Features: LLM-Guided Blockwise Feature Engineering via Executable Program Search Presents a framework that uses language models offline to generate executable feature-engineering code for tabular risk models. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Addresses industrial risk-control modeling rather than deployment safety, jailbreaks, or alignment.
Long Text to Predictive Features: LLM-Guided Blockwise Feature Engineering via Executable Program Search Presents a framework that uses language models offline to generate executable feature-engineering code for tabular risk models. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Focuses on program synthesis for feature extraction rather than model compression or efficient serving.
Long Text to Predictive Features: LLM-Guided Blockwise Feature Engineering via Executable Program Search Presents a framework that uses language models offline to generate executable feature-engineering code for tabular risk models. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Generates offline feature-engineering programs from text without performing retrieval-augmented generation.
SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models Introduces a benchmark evaluating how well vision-language models predict unseen physical outcomes following scene interventions. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Focuses on multimodal benchmark evaluation rather than model compression, quantization, or efficient serving.
SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models Introduces a benchmark evaluating how well vision-language models predict unseen physical outcomes following scene interventions. Preprint	LLMs for code via arxiv · 12 Oct, 16:28	Not relevant Does not address code generation, program repair, or software engineering tasks.
SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models Introduces a benchmark evaluating how well vision-language models predict unseen physical outcomes following scene interventions. Preprint	AI agents via arxiv · 12 Oct, 16:28	Not relevant Evaluates predictive spatial reasoning in vision-language models, not multi-step tool-using autonomous agents.
SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models Introduces a benchmark evaluating how well vision-language models predict unseen physical outcomes following scene interventions. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Evaluates spatial reasoning accuracy rather than AI safety, guardrails, or jailbreaks.
SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models Introduces a benchmark evaluating how well vision-language models predict unseen physical outcomes following scene interventions. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Focuses on spatial reasoning in vision-language models rather than retrieval-augmented generation systems.
SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models Introduces a benchmark evaluating how well vision-language models predict unseen physical outcomes following scene interventions. Preprint	Computer vision for industrial applications via arxiv · 12 Oct, 16:28	Not relevant Benchmark covers general 3D scenes rather than industrial inspection or manufacturing computer vision.
ViSkill: Reinforcing VLM Agents with Evolving Visual-Native Skills Proposes a framework that encodes visual trajectories into reusable skill cards to reinforce vision-language agent planning and rewards. Preprint	LLMs for code via arxiv · 12 Oct, 16:28	Not relevant Evaluates visual RL environments rather than programming or software code.
ViSkill: Reinforcing VLM Agents with Evolving Visual-Native Skills Proposes a framework that encodes visual trajectories into reusable skill cards to reinforce vision-language agent planning and rewards. Preprint	Computer vision for industrial applications via arxiv · 12 Oct, 16:28	Not relevant Tests on abstract game environments (e.g. Sokoban) rather than industrial vision inspection.
ViSkill: Reinforcing VLM Agents with Evolving Visual-Native Skills Proposes a framework that encodes visual trajectories into reusable skill cards to reinforce vision-language agent planning and rewards. Preprint	LLM evaluation and benchmarks via arxiv · 12 Oct, 16:28	Not relevant Proposes an agent framework and training method rather than an evaluation benchmark.
ViSkill: Reinforcing VLM Agents with Evolving Visual-Native Skills Proposes a framework that encodes visual trajectories into reusable skill cards to reinforce vision-language agent planning and rewards. Preprint	Small and efficient models via arxiv · 12 Oct, 16:28	Not relevant Focuses on reinforcement learning sample efficiency rather than model compression or inference latency.
ViSkill: Reinforcing VLM Agents with Evolving Visual-Native Skills Proposes a framework that encodes visual trajectories into reusable skill cards to reinforce vision-language agent planning and rewards. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Focuses on task completion performance rather than safety guardrails or alignment.
ViSkill: Reinforcing VLM Agents with Evolving Visual-Native Skills Proposes a framework that encodes visual trajectories into reusable skill cards to reinforce vision-language agent planning and rewards. Preprint	Retrieval-augmented generation (RAG) via arxiv · 12 Oct, 16:28	Not relevant Focuses on reinforcement learning policy skills rather than document retrieval or RAG.
WOVEN: Weaving Visual World Modeling into Multimodal LLMs Presents a benchmark and dataset using generative video rollouts to train and evaluate visual transition reasoning in multimodal models. Preprint	AI safety and reliability in deployment via arxiv · 12 Oct, 16:28	Not relevant Investigates general spatial and physical reasoning rather than deployment safety or guardrails.

12,301–12,360 of 13,754← Previous Next →