{"arxiv_id":"2608.07449","incoming":{"papers":0,"sentences":0,"groups":[],"unlabelled_sentences":0},"outgoing":{"papers":21,"sentences":27,"groups":[{"kind":"data_used","label":"Uses data or values from","papers":3,"sentences":3,"quotes":[{"arxiv_id":"2406.14991","title":"SpreadsheetBench: Towards Challenging Real World Spreadsheet Manipulation","url":"/paper/2406.14991","section":"Experimental Setups","sentence":"For the IID evaluation, we adopt SpreadsheetBench Verified [ref] , a human-validated subset of SpreadsheetBench that is specifically curated to enable reliable automatic evaluation.","intent_p":0.91,"contests_p":null},{"arxiv_id":"1508.00305","title":"Compositional Semantic Parsing on Semi-Structured Tables","url":"/paper/1508.00305","section":"Experimental Setups","sentence":"For the OOD evaluation, we further evaluate our method on WikiTableQuestions [ref] and HiTab [ref] , which allows us to examine the generalization ability of the learned skills beyond the training distribution.","intent_p":0.99,"contests_p":null},{"arxiv_id":"2108.06712","title":"HiTab: A Hierarchical Table Dataset for Question Answering and Natural Language Generation","url":"/paper/2108.06712","section":"Experimental Setups","sentence":"For the OOD evaluation, we further evaluate our method on WikiTableQuestions [ref] and HiTab [ref] , which allows us to examine the generalization ability of the learned skills beyond the training distribution.","intent_p":0.99,"contests_p":null}]},{"kind":"comparison","label":"Compares with","papers":4,"sentences":4,"quotes":[{"arxiv_id":"2605.23904","title":"SkillOpt: Executive Strategy for Self-Evolving Agent Skills","url":"/paper/2605.23904","section":"Experimental Setups","sentence":"Third, for methods that focus on skill self-evolution, we compare against SkillGrad [ref] and SkillOpt [ref] .","intent_p":1.0,"contests_p":null},{"arxiv_id":"2605.27760","title":"SkillGrad: Optimizing Agent Skills Like Gradient Descent","url":"/paper/2605.27760","section":"Experimental Setups","sentence":"Third, for methods that focus on skill self-evolution, we compare against SkillGrad [ref] and SkillOpt [ref] .","intent_p":1.0,"contests_p":null},{"arxiv_id":"2603.02766","title":"EvoSkill: Automated Skill Discovery for Multi-Agent Systems","url":"/paper/2603.02766","section":"Experimental Setups","sentence":"Second, for methods that leverage LLM-generated skills, we compare against EvoSkill [ref] and Trace2Skill [ref] .","intent_p":1.0,"contests_p":null},{"arxiv_id":"2603.25158","title":"Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills","url":"/paper/2603.25158","section":"Experimental Setups","sentence":"Second, for methods that leverage LLM-generated skills, we compare against EvoSkill [ref] and Trace2Skill [ref] .","intent_p":1.0,"contests_p":null}]},{"kind":"background","label":"Cites as background","papers":7,"sentences":7,"quotes":[{"arxiv_id":"2605.23904","title":"SkillOpt: Executive Strategy for Self-Evolving Agent Skills","url":"/paper/2605.23904","section":"Related Works","sentence":"SkillOpt [ref] bounds each edit with a textual learning rate and a held-out gate, while SkillGrad [ref] treats the skill package as a structured parameter updated by trajectory-level loss evidence, textual gradients, and momentum.","intent_p":0.86,"contests_p":null},{"arxiv_id":"2605.27760","title":"SkillGrad: Optimizing Agent Skills Like Gradient Descent","url":"/paper/2605.27760","section":"Related Works","sentence":"SkillOpt [ref] bounds each edit with a textual learning rate and a held-out gate, while SkillGrad [ref] treats the skill package as a structured parameter updated by trajectory-level loss evidence, textual gradients, and momentum.","intent_p":0.86,"contests_p":null},{"arxiv_id":"2602.08234","title":"SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning","url":"/paper/2602.08234","section":"Related Works","sentence":"Recent LLM agents increasingly externalize procedural knowledge into skills—structured bundles of instructions, scripts, and resources that are loaded into the agent's context at inference time and improve task execution without any weight update [ref] .","intent_p":0.87,"contests_p":null},{"arxiv_id":"2602.12430","title":"Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward","url":"/paper/2602.12430","section":"Related Works","sentence":"Recent LLM agents increasingly externalize procedural knowledge into skills—structured bundles of instructions, scripts, and resources that are loaded into the agent's context at inference time and improve task execution without any weight update [ref] .","intent_p":0.87,"contests_p":null},{"arxiv_id":"2604.08224","title":"Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering","url":"/paper/2604.08224","section":"Related Works","sentence":"Recent LLM agents increasingly externalize procedural knowledge into skills—structured bundles of instructions, scripts, and resources that are loaded into the agent's context at inference time and improve task execution without any weight update [ref] .","intent_p":0.87,"contests_p":null}]}],"unlabelled_sentences":13},"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/2608.07449"}}