NVIDIA、生物基盤モデル用LoRAレシピ公開

Native Active Perception as Reasoning for Omni-Modal Understanding

arXiv cs.CL (Computation and Language) ・ 2026-06-17

動画理解を効率化、難易度に応じ能動的に知覚する推論手法を提案

Does VLA Even Know the Basics? Measuring Commonsense and World Knowledge Retention in Vision-Language-Action Models

arXiv cs.LG (Machine Learning) ・ 2026-06-17

VLA モデルは常識を保持しているか、知識保持度を測る研究

Trade-offs in Medical LLM Adaptation: An Empirical Study in French QA

arXiv cs.AI (Artificial Intelligence) ・ 2026-06-17

医療 LLM のドメイン適応、仏語 QA で利点と代償を実証研究

Mechanism-Guided Selective Unlearning for RLVR-Induced Reasoning

arXiv cs.AI (Artificial Intelligence) ・ 2026-06-17

MAST、RLVR 由来の推論を巻き添えを抑えて選択的に忘却

Beyond Safe Data: Pretraining-Stage Alignment with Regular Safety Reflection

arXiv cs.AI (Artificial Intelligence) ・ 2026-06-17

事前学習段階で安全反映、安全データを超えるアライメントを探る

ProductConsistency: Improving Product Identity Preservation in Instruction-Based Image Editing via SFT and RL

arXiv cs.AI (Artificial Intelligence) ・ 2026-06-17

ProductConsistency、指示画像編集で商品の同一性保持を改善

ARIADNE: Agnostic Routing for Inference-time Adapter DyNamic sElection

arXiv cs.AI (Artificial Intelligence) ・ 2026-06-17

ARIADNE、推論時にアダプタを動的選択する汎用ルーティング

Learning from the Self-future: On-policy Self-distillation for dLLMs

arXiv cs.CL (Computation and Language) ・ 2026-06-16

拡散LLM向けのオンポリシー自己蒸留OPSDを探究

Multi-Source Cybersecurity Logs: An ATT&CK-Labeled Dataset and SLM Evaluation

arXiv cs.LG (Machine Learning) ・ 2026-06-16

ATT&CK注釈付き多源セキュリティログ集と小型LM評価

S4oP: Operator-level Pruning of Structured State Space Models for Resource-Constrained Devices

arXiv cs.LG (Machine Learning) ・ 2026-06-16

S4oP、状態空間モデルを演算子単位で枝刈り軽量化

EAGG: Embodiment-Aligned Grasp Generation via Geometry-Aware Graph Conditioning

arXiv cs.AI (Artificial Intelligence) ・ 2026-06-16

身体性に整合した把持生成手法EAGGを提案

From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoning

arXiv cs.LG (Machine Learning) ・ 2026-06-16

推論トレースを再利用モジュール化し合成的汎化を解明

Uncertainty Quantification for Flow-Based Vision-Language-Action Models

arXiv cs.LG (Machine Learning) ・ 2026-06-16

フローベースVLAモデルの不確実性定量化

When English Isn't the Best Teacher: Source Language Effects in Cross-Lingual In-Context Learning

arXiv cs.CL (Computation and Language) ・ 2026-06-16

言語間文脈内学習で英語が最良の教師とは限らない

Catastrophic Forgetting is Low-Rank: A Function-Space Theory for Continual Adaptation

arXiv cs.LG (Machine Learning) ・ 2026-06-16

破滅的忘却は低ランク、継続適応の関数空間理論

Fine-tuning LLMs for Passive Depression Severity Estimation from AI Mental Health Dialogue

arXiv cs.CL (Computation and Language) ・ 2026-06-16

AIメンタルヘルス対話からうつ重症度を受動推定するLLM微調整

Environment-Grounded Automated Prompt Optimization for LLM Game Agents

arXiv cs.CL (Computation and Language) ・ 2026-06-16

環境に接地した自動プロンプト最適化でLLMゲームエージェント

From Drift to Coherence: Stabilizing Beliefs in LLMs

arXiv cs.LG (Machine Learning) ・ 2026-06-16

ドリフトから整合へ、LLMの信念を安定化

Improving low-resource ASR using bilingual fine-tuning with language identification: a cross-linguistic evaluation

arXiv cs.CL (Computation and Language) ・ 2026-06-16

言語識別付き二言語微調整で低資源ASRを改善

SuCo: Sufficiency-guided Continuous Adaptive Reasoning

arXiv cs.CL (Computation and Language) ・ 2026-06-16

SuCo、十分性に導かれた連続適応的推論

The Value Axis: Language Models Encode Whether They're on the Right Track

arXiv cs.CL (Computation and Language) ・ 2026-06-15

LLM内部に戦略の成否を符号化する「価値軸」を発見

Hierarchical Advantage Weighting for Online RL Fine-Tuning of VLAs from Sparse Episode Outcomes

arXiv cs.LG (Machine Learning) ・ 2026-06-15

VLAのオンラインRL微調整に階層的優位重み付けHABCを提案

KVEraser: Learning to Steer KV Cache for Efficient Localized Context Erasing

arXiv cs.CL (Computation and Language) ・ 2026-06-15

KVキャッシュの局所消去を学習で効率化するKVEraserを提案

ExpRL: Exploratory RL for LLM Mid-Training

arXiv cs.LG (Machine Learning) ・ 2026-06-15

ExpRL、人手QAデータを「報酬足場」に使うLLM中間学習向けRLを提案

Selection Without Signal, Recovery Through Expression: A Measurement Study of Post-Hoc Falsification Operators for Frozen Small Code Models

arXiv cs.CL (Computation and Language) ・ 2026-06-15

凍結された小型コードモデル向け事後反証演算子の測定的研究

Exploring Extrinsic and Intrinsic Properties for Effective Reasoning with Code Interpreter

arXiv cs.CL (Computation and Language) ・ 2026-06-15

コードインタープリタ推論を支える内在・外在特性を分析した論文

Robust Dual-Signal Fusion: Hybrid Neuro-Symbolic Gating with Compressed Chain-of-Thought Refinement for Irony Detection in Social Media Texts

arXiv cs.CL (Computation and Language) ・ 2026-06-15

圧縮 CoT の神経記号ハイブリッドでゼロショット皮肉検出を強化

ATOM-Bench: A Real-World Benchmark for Atomic Skills and Compositional Generalization in Manipulation Policies

arXiv cs.AI (Artificial Intelligence) ・ 2026-06-15

ロボット操作方策の原子的技能と合成的汎化を測るATOM-Benchを提案

Adaptive and Explicit safe: Triggering Latent Safety Awareness in Large Reasoning Models

arXiv cs.AI (Artificial Intelligence) ・ 2026-06-15

推論モデルの潜在的な安全認識を引き出し脱獄耐性を高める手法を提案

The Art of Mixology: Mixup-based Obfuscation for Privacy-Preserving Split Learning in Large Language Models

arXiv cs.CL (Computation and Language) ・ 2026-06-15

LLMのスプリット学習向けプライバシー保護手法MIXGUARDを提案

daVinci-kernel: Co-Evolving Skill Selection, Summarization, and Utilization via RL for GPU Kernel Optimization

arXiv cs.CL (Computation and Language) ・ 2026-06-15

GPUカーネル最適化向けスキル共進化RL「daVinci-kernel」を提案