{"arxiv_id":"2608.08303","incoming":{"papers":0,"sentences":0,"groups":[],"unlabelled_sentences":0},"outgoing":{"papers":28,"sentences":45,"groups":[{"kind":"method_or_tool","label":"Uses the method or tool of","papers":5,"sentences":5,"quotes":[{"arxiv_id":"2603.25158","title":"Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills","url":"/paper/2603.25158","section":"Experimental Settings","sentence":"We evaluate attack methods on SkillOpt [ref] and Trace2Skill [ref] , which represent performance-guided iterative skill optimization and trajectory-level skill consolidation, respectively.","intent_p":0.91,"contests_p":null},{"arxiv_id":"2605.23904","title":"SkillOpt: Executive Strategy for Self-Evolving Agent Skills","url":"/paper/2605.23904","section":"Experimental Settings","sentence":"We evaluate attack methods on SkillOpt [ref] and Trace2Skill [ref] , which represent performance-guided iterative skill optimization and trajectory-level skill consolidation, respectively.","intent_p":0.91,"contests_p":null},{"arxiv_id":"2508.10925","title":"gpt-oss-120b & gpt-oss-20b Model Card","url":"/paper/2508.10925","section":"Experimental Settings","sentence":"We use GPT-OSS-20B and GPT-OSS-120B [ref] , together with their corresponding safeguard variants [ref] for the defense evaluation.","intent_p":0.93,"contests_p":null},{"arxiv_id":"2602.02276","title":"Kimi K2.5: Visual Agentic Intelligence","url":"/paper/2602.02276","section":"Experimental Settings","sentence":"To assess generalization across different backbone models, we additionally evaluate TBA using two proprietary models, GPT-5.4 [ref] and GPT-5.5 [ref] , and two open-weight models, Kimi-K2.5 [ref] , Qwen3-VL-235B-A22B [ref] .","intent_p":0.84,"contests_p":null},{"arxiv_id":"2511.21631","title":"Qwen3-VL Technical Report","url":"/paper/2511.21631","section":"Experimental Settings","sentence":"To assess generalization across different backbone models, we additionally evaluate TBA using two proprietary models, GPT-5.4 [ref] and GPT-5.5 [ref] , and two open-weight models, Kimi-K2.5 [ref] , Qwen3-VL-235B-A22B [ref] .","intent_p":0.84,"contests_p":null}]},{"kind":"supports_this_paper","label":"Cites as support","papers":2,"sentences":2,"quotes":[{"arxiv_id":"2605.23904","title":"SkillOpt: Executive Strategy for Self-Evolving Agent Skills","url":"/paper/2605.23904","section":"Introduction","sentence":"Self-evolving systems therefore shift skill acquisition from external artifacts to trusted internal construction and have demonstrated improved performance and adaptability over static skills [ref] .","intent_p":0.98,"contests_p":null},{"arxiv_id":"2311.09277","title":"Contrastive Chain-of-Thought Prompting","url":"/paper/2311.09277","section":"Counterfactual Trajectory Induction","sentence":"Prior work shows that jointly presenting valid and invalid behaviors helps language models distinguish a desired pattern from plausible alternatives [ref] .","intent_p":0.96,"contests_p":null}]},{"kind":"background","label":"Cites as background","papers":13,"sentences":13,"quotes":[{"arxiv_id":"2605.23904","title":"SkillOpt: Executive Strategy for Self-Evolving Agent Skills","url":"/paper/2605.23904","section":"Self-Evolving Skill Systems","sentence":"SkillOpt [ref] optimizes a skill document through rollout-based revisions selected by held-out evaluation.","intent_p":0.88,"contests_p":null},{"arxiv_id":"2603.01145","title":null,"url":"/paper/2603.01145","section":"Self-Evolving Skill Systems","sentence":"Voyager [ref] stores executable skills that pass environment-based self-verification, whereas AutoSkill [ref] continually structures accumulated experience into reusable knowledge for lifelong adaptation.","intent_p":0.95,"contests_p":null},{"arxiv_id":"2601.10338","title":"Agent Skills in the Wild: An Empirical Study of Security Vulnerabilities at Scale","url":"/paper/2601.10338","section":"Attacks on Self-Evolving Agents","sentence":"Self-evolving agents improve after deployment by learning from interaction trajectories, environmental feedback, and self-reflection [ref] .","intent_p":0.99,"contests_p":null},{"arxiv_id":"2507.21046","title":"A Survey of Self-Evolving Agents: What, When, How, and Where to Evolve on the Path to Artificial Super Intelligence","url":"/paper/2507.21046","section":"Attacks on Self-Evolving Agents","sentence":"Self-evolving agents improve after deployment by learning from interaction trajectories, environmental feedback, and self-reflection [ref] .","intent_p":0.99,"contests_p":null},{"arxiv_id":"2508.07407","title":"A Comprehensive Survey of Self-Evolving AI Agents: A New Paradigm Bridging Foundation Models and Lifelong Agentic Systems","url":"/paper/2508.07407","section":"Attacks on Self-Evolving Agents","sentence":"Self-evolving agents improve after deployment by learning from interaction trajectories, environmental feedback, and self-reflection [ref] .","intent_p":0.99,"contests_p":null}]}],"unlabelled_sentences":25},"claims":[],"thresholds":{"intent_p":0.8,"supports_p":0.95,"contests_p":0.7},"method":"Citing sentences come from each paper's LaTeX source; every \\cite marks the sentence and the works it cites. Pith's classifier typed each sentence. A label is shown at intent probability 0.8 or more (0.95 for support), a contest at 0.7 or more. Measured accuracy against two blind model families is at https://pith.science/accuracy. The quoted sentences are the citing authors' own.","held_up":null,"_links":{"html":"https://pith.science/paper/2608.08303"}}