{"paper":{"title":"SEAR: Self-Evolving Agentic Robotics","arxiv_id":"2610.18792","primary_cat":"cs.RO","submitted_at":"2026-10-13T17:40:20+00:00","version_compiled":"v1"},"notice":"Every quoted sentence is the authors' text from the arXiv listing abstract. Labels and probabilities are the outputs of classifier-1.13.0 on the question bank abstract-p1-20260923, exactly as returned. No sentence in this record was written by a language model.","schema":"pith.compiled-record/v0","status":"compiled","abstract":{"source":"arXiv listing abstract v1","sentences":[{"i":0,"p":1.0,"id":"s0","own":false,"role":"background","text":"Frontier language-model agents can now improve robots on their own: they write control code, test it, and keep what works as skills and memories for the next task.","hedge":{"p":0.91,"label":"asserted"},"own_p":0.11,"main_claim_p":0.05,"quantitative":false,"quantitative_p":0.05,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":1.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":1,"p":1.0,"id":"s1","own":true,"role":"contribution","text":"We introduce SEAR (Self-Evolving Agentic Robotics), a framework and benchmark built around two questions.","hedge":{"p":0.89,"label":"asserted"},"own_p":0.84,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.44,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":1.0,"interpretation":0.0}},{"i":2,"p":0.64,"id":"s2","own":false,"role":"gap","text":"Is the agent really learning, or just trying more often?","hedge":{"p":0.95,"label":"unclear"},"own_p":0.08,"main_claim_p":0.02,"quantitative":false,"quantitative_p":0.05,"role_probabilities":{"gap":0.64,"method":0.01,"result":0.0,"background":0.07,"limitation":0.0,"future_work":0.0,"contribution":0.28,"interpretation":0.0}},{"i":3,"p":0.52,"id":"s3","own":false,"role":"gap","text":"And does what it learns still help on new tasks, in a new simulator, and on a real robot?","hedge":{"p":0.95,"label":"unclear"},"own_p":0.15,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.05,"role_probabilities":{"gap":0.52,"method":0.01,"result":0.0,"background":0.12,"limitation":0.0,"future_work":0.0,"contribution":0.35,"interpretation":0.0}},{"i":4,"p":0.96,"id":"s4","own":true,"role":"method","text":"The framework runs an agent in three modes: it writes code that becomes the policy (code-as-policy), acts live in the control loop (agent-as-policy), or works as a researcher that fine-tunes a vision-language-action model (auto-research).","hedge":{"p":1.0,"label":"asserted"},"own_p":0.76,"main_claim_p":0.03,"quantitative":true,"quantitative_p":0.56,"role_probabilities":{"gap":0.0,"method":0.96,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.04,"interpretation":0.0}},{"i":5,"p":0.79,"id":"s5","own":true,"role":"method","text":"The benchmark adds controls that separate retained experience from repeated attempts, familiar tasks, privileged simulator state, larger budgets, and luck.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.84,"main_claim_p":0.04,"quantitative":false,"quantitative_p":0.14,"role_probabilities":{"gap":0.0,"method":0.79,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.21,"interpretation":0.0}},{"i":6,"p":0.48,"id":"s6","own":true,"role":"method","text":"Across seven frontier language models, 330 simulated and five physical tasks, and over four years of agent runtime, we highlight three findings.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.96,"main_claim_p":0.05,"quantitative":true,"quantitative_p":0.98,"role_probabilities":{"gap":0.0,"method":0.48,"result":0.14,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.38,"interpretation":0.0}},{"i":7,"p":1.0,"id":"s7","own":true,"role":"result","text":"First, the learning curve of a robotic agent depends both on the overall time budget and the task difficulty relative to model capability.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.93,"main_claim_p":0.23,"quantitative":false,"quantitative_p":0.13,"role_probabilities":{"gap":0.0,"method":0.0,"result":1.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":8,"p":0.8,"id":"s8","own":true,"role":"result","text":"Harder tasks make longer sessions worthwhile; on easier tasks, re-rolling short attempts does just as well.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.93,"main_claim_p":0.13,"quantitative":true,"quantitative_p":0.73,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.8,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.2}},{"i":9,"p":1.0,"id":"s9","own":true,"role":"result","text":"Second, control code and shared routines transfer across tasks and simulators, while task-specific solutions do not.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.92,"main_claim_p":0.21,"quantitative":false,"quantitative_p":0.36,"role_probabilities":{"gap":0.0,"method":0.0,"result":1.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":10,"p":0.96,"id":"s10","own":true,"role":"result","text":"On a real robot, we observe a performance gap between agent-designed practice and novel human-designed scenes.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.95,"main_claim_p":0.07,"quantitative":true,"quantitative_p":0.78,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.96,"background":0.0,"limitation":0.04,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":11,"p":1.0,"id":"s11","own":true,"role":"result","text":"Third, different models evolve more effectively in different ways.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.92,"main_claim_p":0.13,"quantitative":false,"quantitative_p":0.22,"role_probabilities":{"gap":0.0,"method":0.0,"result":1.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":12,"p":0.99,"id":"s12","own":true,"role":"result","text":"Some do best acting live, others writing code first.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.9,"main_claim_p":0.08,"quantitative":false,"quantitative_p":0.32,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.99,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.01}},{"i":13,"p":0.8,"id":"s13","own":true,"role":"result","text":"And a text-only model keeps pace with a comparable multimodal one: perception helps with error assessment, but agentic reasoning and tool use can matter more.","hedge":{"p":0.91,"label":"asserted"},"own_p":0.92,"main_claim_p":0.1,"quantitative":true,"quantitative_p":0.75,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.8,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.2}},{"i":14,"p":0.8,"id":"s14","own":true,"role":"contribution","text":"We release the benchmark and codebase to support future research.","hedge":{"p":0.91,"label":"asserted"},"own_p":0.85,"main_claim_p":0.02,"quantitative":false,"quantitative_p":0.02,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.2,"contribution":0.8,"interpretation":0.0}}]},"compiled":{"main_claim":{"p":0.28,"text":"Frontier language-model agents can now improve robots on their own: they write control code, test it, and keep what works as skills and memories for the next task.","location":"abstract s0","hedge_strength":{"p":0.91,"label":"asserted"}},"provenance":{"route":{"via":"TypeSafe API","model":"classifier-1.13.0","endpoint":"https://api.typesafe.ai/v1/systemone","provider":"TypeSafe"},"packing":"one request per paper","requests":1,"input_tokens":10628,"cost_usd_reported":0.0},"compiled_at":"2026-10-14","bank_version":"abstract-p1-20260923","model_version":"classifier-1.13.0","abstract_states":{"claims_first":0.07,"states_limitation":0.84,"human_participants":0.07,"controlled_comparison":0.79,"randomized_assignment":0.02,"states_quantitative_result":0.89,"compares_with_prior_results":0.5,"mentions_code_or_data_availability":0.98},"evidential_basis":{"p":0.94,"value":"observation or experiment","probabilities":{"argument":0.0,"literature":0.0,"proof_or_derivation":0.0,"analysis_of_existing_data":0.0,"observation_or_experiment":0.94,"simulation_or_computation":0.06}},"contribution_type":{"p":0.84,"value":"dataset or benchmark","probabilities":{"review":0.0,"method_or_tool":0.12,"empirical_study":0.04,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":0.84,"position_or_commentary":0.0}},"contribution_statement":{"p":0.99,"text":"We introduce SEAR (Self-Evolving Agentic Robotics), a framework and benchmark built around two questions.","location":"abstract s1"}},"coverage":{"tier":"abstract","levels_run":["P abstract"],"not_examined":["full text","figures and tables","references and citing sentences","relations to other papers"],"sources_read":["arXiv listing abstract v1"],"levels_not_run":["S sections","T body sentences","C citations","R cross-paper relations"],"sentences_examined":15,"sentences_in_abstract":15},"measured_accuracy":{"url":"https://pith.science/accuracy","measured_on":"2026-10-02","method":"two independent model families labelled a fresh random sample blind; the figure is how often the published label matches their consensus","sentence_role_agreement":0.92,"main_claim_agreement":0.908,"typed_citation_agreement":0.933,"typed_citation_agreement_95_interval":[0.917,0.945],"review_comments_accurate":0.758,"review_comments_accurate_95_interval":[0.684,0.82],"review_caught_admitted_error_found_or_related":0.448,"review_comments_judged":{"judged_by":["gpt-5.6-luna","grok-4.7"],"judged_against":"Pith's own text copy of each paper: text extracted from the PDF, the first 90,000 characters","measured_on":"2026-10-04","accurate":113,"decided":149,"sampled_reviews":200,"split_by_lane":false,"by_lane":"https://pith.science/accuracy#by-lane"}}}