Tag
SkillMem introduces a compact external latent memory for reusable skills in latent multi-agent reasoning, achieving 38× storage reduction and up to 4.23× inference speedup across benchmarks with Qwen3 models from 4B to 14B.
SkillShift-Bench introduces the first benchmark for evaluating web agent skill learning under controlled open-environment shifts (contextual, perceptual, and execution), revealing systematic robustness degradation and proposing a test-time adaptation method to improve shifted skill execution without relearning.
该论文提出Δτ-Bench，一个评估AI智能体在不断演化的工具生态中适应能力的可执行基准，包含42个领域、6个生态状态下的1,588个多轮任务；并提出Steward生态系统管理器，通过增量技能维护和动态供应实现更高的持续任务成功率。
SkillNEO 提出非精英进化优化方法来改进 LLM agent 技能，通过允许低奖励版本继续迭代并分析版本间差异来保留有效步骤、修复失败，使 GPT-5.4 Mini 在 86 项 SkillsBench 任务上平均奖励从 49.1% 提升至 66.4%，且改进后的技能还能提升更强模型的表现。
EmbodiedEvo 让编码代理在模拟机器人场景中编写 Python 代码、执行任务并将经验固化为可复用的技能，从而零样本迁移到真实机器人与新环境，无需更新模型权重。
提出了 SkillGrain 框架，将用户反馈分解为可独立执行的原子规则，预测其可迁移性并只保留适用于未来请求的部分，通过强化学习联合优化技能归纳与条件生成，以实现可复用的个性化技能学习。
该论文提出 Socratic Skill Learning (SSL)，通过将专家教学与学生实践交错进行，使 LLM 代理获得可迁移的可复用技能，并构建了 SkillTransferBench 基准测试，结果显示其在效用指标上优于轨迹蒸馏与对比蒸馏方法。
UC Berkeley 与 Princeton 研究者提出一种可扩展的 empowerment（赋能）最大化配方，将赋能与可执行技能数量联系起来，并在 Crafter 等复杂环境中展示出建房、资源收集等涌现行为。
提出了RSIPO框架，通过技能与策略的协同演化实现智能体强化学习的递归自我优化，在ALFWorld上以Qwen2.5-7B达到98.6%的成功率，显著超越GRPO等基线。
ViSkill 提出一种视觉原生的技能学习框架，将成功交互编码为可复用的视觉技能卡，与策略优化形成闭环反馈，在 Sokoban、FrozenLake 等基准上超越多种基线。
UniSkill 提出一种共享策略与技能库共同演化的框架，通过对比动作反馈作为技能提案的 actor-alignment 信号，避免昂贵的额外 rollout，在 ALFWorld 达到 98.4% 成功率、WebShop 达到 84.7%。
Skill2Env is a capability-oriented framework that synthesizes executable environments from skills to train general AI agents, using task blueprints and iterative hardening to improve agent performance across benchmarks.
Skild AI demonstrated a robot trained to play football using self-play simulation equivalent to 140 years of practice in weeks, illustrating a new approach to skill acquisition through computer simulation.
EvoSkill v2 enables AI agents to self-improve without retraining by using persistent skills stored in files, with a coach agent writing lessons from failed tasks, leading to significant performance gains in tasks like spreadsheet repair.
GeoSkill introduces a hierarchical skill learning framework for geospatial agents to distill execution experience into reusable skills, enhancing task accuracy and tool reliability through collaborative revision.
SPACE 通过从成功轨迹归纳两层参数化技能，将子技能边界作为动作块监督，训练长程 LLM Agent 自适应输出可变长原子动作序列；在 ALFWorld 和 ScienceWorld 上成功率提升 7.0%–31.3%，决策轮次最多降低 78.9%。
Introduces QDOS, a unified pipeline for offline-to-online reinforcement learning that uses advantage-weighted quality-diversity pretraining to extract diverse and high-value skills, significantly improving performance in manipulation and locomotion tasks.
This article presents a method to extract ideas from bloggers' content using AI tools and develop reusable skill workflows.
SAGE is a framework for automating storyboard generation in short drama production using self-evolving rules and attribution-guided updates, achieving expert-level performance and reducing authoring time in commercial deployment.
Introduces ERSkill, a retrieval-centric framework for self-evolving, skill-guided adaptive memory access in LLM agents. It co-evolves retrieval skills and a routing policy, substantially outperforming strong baselines across agent memory benchmarks.