📄 arXiv 论文速递

💡 LLM评估器被广泛用作评分工具和训练信号，但其内部机制尚不明确。本文揭示其工作原理，有助于提升评估可靠性和可解释性，对NLG评测领域具有基础性价值。

LLM-based evaluators of natural language generation (NLG) quality are widely deployed as scoring tools and as automated training signals, yet the internal procedure by which they a…

💡 软件工程智能体评测成本高昂，每个任务需多步探索和修改。本文提出轨迹感知的高效评测方法，大幅降低基准测试开销，加速智能体研发迭代，对实际部署至关重要。

Evaluating software engineering agents on realistic benchmarks is costly, since each task may require multi-step code exploration, modification, and test execution. Existing effici…

💡 仓库级代码生成需兼顾任务需求与仓库一致性，现有检索方法常忽略关键token。本文提出自适应关键token感知检索，提升代码生成准确性与仓库适配度，推动自动化开发进步。

The repository-level code generation task requires synthesizing code that satisfies task requirements while remaining consistent with the target repository context. Since real-worl…

💡 动态智能体框架允许语言模型改变自身执行环境，带来新的推理负担。本文提出基准测试评估模型对组件生命周期的推理能力，揭示关键局限，对构建安全自适应智能体至关重要。

Dynamic agent harnesses let language models change the software that shapes their own execution. This flexibility brings a new reasoning burden: a local plugin change can propagate…

Natural language is emerging as a primary feedback channel for improving language agents, capable of conveying intent, preferences, and causal structure in forms interpretable by b…

Real-world robotic assembly at sub-millimeter tolerances demands spatial precision, compliant interaction, and robustness to contact failures. We present Facet-0, a robotic foundat…

We develop a framework for mechanism design with AI agents whose alignment (preferences) and capabilities (feasible actions and information) are unknown. We want such agents to act…

Writing involves diverse cognitive activities, from ideation to revision, and writers' needs vary across individuals and moments. Proactive AI promises to provide the right support…

💡 后训练量化广泛用于降低LLM服务成本，但精度损失分布不均。本文揭示量化损伤的结构性特征，提出全局分配比特策略，可显著改善量化模型性能，对高效部署意义重大。

Post-training quantization (PTQ) is widely used to reduce the cost of serving large language models (LLMs), but its accuracy cost is uneven and is often tuned per model. We study w…

How to divide a fixed annotation budget between supervised fine-tuning (SFT) and reinforcement learning (RL) during LLM post-training remains an open problem. Existing work charact…

Vision-Language Models (VLMs) provide useful priors for interactive decision-making, but using them directly as policies is expensive and brittle: they must be queried at every ste…

Large language models (LLMs) struggle to classify text into taxonomies with many semantically similar labels, as the distinctions are domain-specific and not captured by pre-traini…