{ "schema_version": 1, "builder_plan": { "scenario": "Your coding assistant makes changes that still need substantial repair before merging.", "change": "Try the technique on held-out issues from a repository with reliable tests.", "baseline": "The same coding agent, repository snapshot, tools, and token budget.", "steps": [ "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.", "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.", "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.", "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test." ], "measure": "Correct patches, regression failures, human repair time, and total cost per accepted change.", "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation." }, "builder_review": null, "arxiv_id": "2609.20715", "title": "Don't Mask the Environment: Observation Supervision Changes How Agents Explore Under RL", "summary": "ActObs replaces the standard SFT convention of computing loss only on agent action tokens by also supervising the environment observation tokens already present in each trajectory, before running GRPO. It adds no data, parameters, sequence tokens, or forward passes; the cost is a modified loss mask in SFT and the measured pass@1 reliability regression at 8B. It can fail when the deployment budget allows only a single sample (pass@1), since the benefit concentrates in pass@k and exploration diversity, and results are demonstrated only on Qwen3 4B/8B with GRPO on terminal and code-editing tasks.", "rationale": "The change is a low-cost SFT loss-mask modification compatible with a single 24 GB GPU, but the reported gains depend on sampling budget, model size, and GRPO post-training, so it merits a workload-specific A/B test rather than unconditional adoption.", "technique": "Observation-token supervision during agentic SFT", "family": "agentes_de_codigo", "recommendation": "testar", "published": "2026-09-17", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.20715", "independent_implementations": 0, "total_implementations": 1, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "ActObs replaces the standard SFT convention of computing loss only on agent action tokens by also supervising the environment observation tokens already present in each trajectory, before running GRPO. It adds no data, parameters, sequence tokens, or forward passes; the cost is a modified loss mask in SFT and the measured pass@1 reliability regression at 8B. It can fail when the deployment budget allows only a single sample (pass@1), since the benefit concentrates in pass@k and exploration diversity, and results are demonstrated only on Qwen3 4B/8B with GRPO on terminal and code-editing tasks.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.20715", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "On Qwen3-4B, GRPO from ActObs achieves higher pass@k at every evaluated sampling budget than action-only SFT on Terminal-Bench 2.0; on Qwen3-8B it trades some pass@1 for +3.4 pp pass@16, and cross-domain transfer to aider-polyglot improves +4.2 pp pass@1 at 4B. All claims are pass@k percentage-point deltas, not multiplicative factors.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.20715", "source_page": null, "source_excerpt": "" } ], "exposure_map": [ { "dimension": "quality", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compute", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "latency", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "operations", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compatibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "security", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "data_and_training", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "reproducibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" } ], "risks": [], "minimum_test": [], "open_questions": [], "independent_tests": [], "lab_reproductions": [], "report_available": false, "validation_tier": "unknown", "evidence_tier": "unknown", "infrastructure_basis": "unknown", "equations": [], "equations_status": "not_fetched", "core_kind": "", "equations_fetched_at": "", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }