{"paper":{"title":"Are Online Skill and Memory Modules Always Worth Their Tokens? A Budget-Constrained Study of Web Agents","arxiv_id":"2606.15017","primary_cat":"cs.CL","submitted_at":"2026-06-12T23:30:14+00:00","version_compiled":"v2"},"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 v2","sentences":[{"i":0,"p":1.0,"id":"s0","own":false,"role":"background","text":"Online web agents often augment a base actor with memory, workflow, or skill modules.","hedge":{"p":0.92,"label":"asserted"},"own_p":0.04,"main_claim_p":0.01,"quantitative":false,"quantitative_p":0.03,"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":0.51,"id":"s1","own":false,"role":"background","text":"These modules can improve performance, but they also consume test-time tokens, a cost rarely reported alongside the actor's inference cost.","hedge":{"p":0.61,"label":"hedged_possibility"},"own_p":0.1,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.1,"role_probabilities":{"gap":0.49,"method":0.0,"result":0.0,"background":0.51,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":2,"p":0.93,"id":"s2","own":true,"role":"contribution","text":"We study online augmentation, where this overhead is paid on every task, and re-evaluate its benefits under a fixed total inference budget.","hedge":{"p":0.87,"label":"asserted"},"own_p":0.77,"main_claim_p":0.04,"quantitative":false,"quantitative_p":0.18,"role_probabilities":{"gap":0.01,"method":0.06,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.93,"interpretation":0.0}},{"i":3,"p":0.95,"id":"s3","own":true,"role":"method","text":"We compare AWM, ASI, and ReasoningBank with a token-matched vanilla baseline that uses the same budget for additional actor steps.","hedge":{"p":0.87,"label":"asserted"},"own_p":0.87,"main_claim_p":0.04,"quantitative":true,"quantitative_p":0.71,"role_probabilities":{"gap":0.0,"method":0.95,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.05,"interpretation":0.0}},{"i":4,"p":1.0,"id":"s4","own":true,"role":"result","text":"Across four WebArena domains and three models, Gemini 3 Flash, GPT-5.4-mini, and Qwen 3.6-27B, the vanilla baseline matches or surpasses all three augmentation methods in aggregate success rate while often using fewer total tokens.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.96,"main_claim_p":0.7,"quantitative":true,"quantitative_p":0.98,"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":5,"p":0.96,"id":"s5","own":true,"role":"result","text":"We observe a similar trend on WorkArena-L1 with Qwen 3.6-27B, indicating that the effect extends to enterprise knowledge-work tasks.","hedge":{"p":0.73,"label":"hedged_possibility"},"own_p":0.94,"main_claim_p":0.06,"quantitative":false,"quantitative_p":0.29,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.96,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.04}},{"i":6,"p":1.0,"id":"s6","own":true,"role":"interpretation","text":"Our results suggest that skills and workflow memory can be useful in specific domains, but their apparent gains often vanish against a budget-matched actor.","hedge":{"p":0.99,"label":"hedged_possibility"},"own_p":0.9,"main_claim_p":0.85,"quantitative":false,"quantitative_p":0.21,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":1.0}},{"i":7,"p":0.49,"id":"s7","own":true,"role":"result","text":"We further show that run-to-run variance materially affects outcomes and should be reported as a core evaluation criterion for online web agents.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.84,"main_claim_p":0.07,"quantitative":false,"quantitative_p":0.06,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.49,"background":0.0,"limitation":0.0,"future_work":0.24,"contribution":0.05,"interpretation":0.22}}]},"compiled":{"main_claim":{"p":0.94,"text":"Our results suggest that skills and workflow memory can be useful in specific domains, but their apparent gains often vanish against a budget-matched actor.","location":"abstract s6","hedge_strength":{"p":0.99,"label":"hedged possibility"}},"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":6269,"cost_usd_reported":0.0},"compiled_at":"2026-09-23","bank_version":"abstract-p1-20260923","model_version":"classifier-1.13.0","abstract_states":{"claims_first":0.05,"states_limitation":0.58,"human_participants":0.03,"controlled_comparison":0.98,"randomized_assignment":0.02,"states_quantitative_result":0.29,"compares_with_prior_results":0.97,"mentions_code_or_data_availability":0.03},"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":1.0,"value":"empirical study","probabilities":{"review":0.0,"method_or_tool":0.0,"empirical_study":1.0,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":0.0,"position_or_commentary":0.0}},"contribution_statement":{"p":0.76,"text":"We study online augmentation, where this overhead is paid on every task, and re-evaluate its benefits under a fixed total inference budget.","location":"abstract s2"}},"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 v2"],"levels_not_run":["S sections","T body sentences","C citations","R cross-paper relations"],"sentences_examined":8,"sentences_in_abstract":8},"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"}}}