{"paper":{"title":"Contain, Conform, Compare (Extended Version): A Red-Teamed Evaluation Kit for Small Coding Agents and the Full Record of Its Experiments","arxiv_id":"2611.16485","primary_cat":"cs.AI","submitted_at":"2026-11-12T08:54:03+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":0.46,"id":"s0","own":true,"role":"contribution","text":"A hosted score in the Gemma 4 Developer Agent competition is one noisy number, so development needs a local evaluator that is fast, safe for agent-written code, and matching the hosted scorer's prompts. drydock-eval evaluates a small coding agent (Gemma 4 E4B) at about 250 runs per hour on one RTX 5090.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.88,"main_claim_p":0.1,"quantitative":true,"quantitative_p":0.98,"role_probabilities":{"gap":0.31,"method":0.11,"result":0.0,"background":0.12,"limitation":0.0,"future_work":0.0,"contribution":0.46,"interpretation":0.0}},{"i":1,"p":0.91,"id":"s1","own":true,"role":"method","text":"Contain runs each job in a fresh container without network and passed 200 scripted escape tests; conform provides eight checks that compare a local stack's requests with the scorer's configuration, down to rendered tokens; compare provides paired statistics that respect repository clusters.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.89,"main_claim_p":0.08,"quantitative":true,"quantitative_p":0.97,"role_probabilities":{"gap":0.0,"method":0.91,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.09,"interpretation":0.0}},{"i":2,"p":0.48,"id":"s2","own":true,"role":"method","text":"AI agents built and attacked the kit; a human owner approved machine use and release.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.52,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.09,"role_probabilities":{"gap":0.0,"method":0.48,"result":0.0,"background":0.23,"limitation":0.09,"future_work":0.0,"contribution":0.2,"interpretation":0.0}},{"i":3,"p":1.0,"id":"s3","own":true,"role":"contribution","text":"This extended version repeats the short paper and adds the full record of the experiments made with and around the kit: 9,660 rows of agent runs, 2,494 grading controls and 72 serving measurements, all released.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.92,"main_claim_p":0.09,"quantitative":true,"quantitative_p":0.98,"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":4,"p":0.92,"id":"s4","own":true,"role":"result","text":"On four L4 GPUs the 31B model decodes 38.1 tokens per second in one stream and E4B 105.8.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.82,"main_claim_p":0.03,"quantitative":true,"quantitative_p":0.99,"role_probabilities":{"gap":0.0,"method":0.08,"result":0.92,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":5,"p":1.0,"id":"s5","own":true,"role":"result","text":"Six configurations of a 31B agent solved between 37 and 44 of 70 development tasks, and no pairwise difference was detected.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.91,"main_claim_p":0.16,"quantitative":true,"quantitative_p":0.99,"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":6,"p":0.99,"id":"s6","own":true,"role":"result","text":"A pre-registered five-arm experiment (231 tasks, 4,620 jobs) measured four prompt divergences: no E4B solve-rate change was detected (largest estimate -2.1 points, 95% interval -4.4 to +0.4).","hedge":{"p":1.0,"label":"asserted"},"own_p":0.95,"main_claim_p":0.19,"quantitative":true,"quantitative_p":0.99,"role_probabilities":{"gap":0.0,"method":0.01,"result":0.99,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":7,"p":1.0,"id":"s7","own":true,"role":"result","text":"Three changes of the E4B agent led their screens by +1.0 to +6.2 points; each was then tested once on new tasks under a frozen pass rule, and none was confirmed (gains -0.35, -0.39 and +4.10 points).","hedge":{"p":1.0,"label":"asserted"},"own_p":0.94,"main_claim_p":0.1,"quantitative":true,"quantitative_p":0.99,"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":1.0,"id":"s8","own":true,"role":"result","text":"The unchanged baseline, run again on the same 128 tasks, moved by 1.17 points, and 65% of these tasks were solved in none of eight runs.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.95,"main_claim_p":0.06,"quantitative":true,"quantitative_p":0.99,"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":9,"p":0.68,"id":"s9","own":true,"role":"limitation","text":"Prompts and agent bundles are withheld; every run-level result is public.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.73,"main_claim_p":0.05,"quantitative":false,"quantitative_p":0.08,"role_probabilities":{"gap":0.0,"method":0.09,"result":0.0,"background":0.0,"limitation":0.68,"future_work":0.0,"contribution":0.23,"interpretation":0.0}}]},"compiled":{"main_claim":{"p":0.39,"text":"A pre-registered five-arm experiment (231 tasks, 4,620 jobs) measured four prompt divergences: no E4B solve-rate change was detected (largest estimate -2.1 points, 95% interval -4.4 to +0.4).","location":"abstract s6","hedge_strength":{"p":1.0,"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":8014,"cost_usd_reported":0.0},"compiled_at":"2026-11-13","bank_version":"abstract-p1-20260923","model_version":"classifier-1.13.0","abstract_states":{"claims_first":0.05,"states_limitation":0.84,"human_participants":0.12,"controlled_comparison":0.97,"randomized_assignment":0.07,"states_quantitative_result":0.99,"compares_with_prior_results":0.85,"mentions_code_or_data_availability":0.84},"evidential_basis":{"p":0.99,"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":0.99,"simulation_or_computation":0.01}},"contribution_type":{"p":0.99,"value":"method or tool","probabilities":{"review":0.0,"method_or_tool":0.99,"empirical_study":0.01,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":0.0,"position_or_commentary":0.0}},"contribution_statement":{"p":0.82,"text":"This extended version repeats the short paper and adds the full record of the experiments made with and around the kit: 9,660 rows of agent runs, 2,494 grading controls and 72 serving measurements, all released.","location":"abstract s3"}},"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":10,"sentences_in_abstract":10},"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"}}}