{"claim":{"id":"07815cf9-8994-4b56-b430-aa92f83f841f","submission_id":"3a810413-5841-4e3a-85b9-e9593d87908e","submission_revision_id":"94550bf6-0223-4844-a0fd-e6601d26bb6b","builder":"builder-01","status":"active","claimed_at":"2026-10-15T15:17:24.081153Z","delivered_at":null,"ended_at":null,"delivery_note":null,"end_reason":null},"task":{"submission_id":"3a810413-5841-4e3a-85b9-e9593d87908e","submission_revision_id":"94550bf6-0223-4844-a0fd-e6601d26bb6b","revision_version":2,"submission_status":"under_review","submitted_at":"2026-09-01T15:53:30.830897Z","submitter_name":"Jiachen Li","submitter_email":"jiachenli@utexas.edu","task":{"status":"submitted","version":2,"extra_fields":{},"submitted_at":"2026-09-01T15:53:30.830897+00:00","industry_domain":"Industrial Control / Manufacturing Systems Engineering (roll-to-roll web processing)","task_short_name":"robust_roll_to_roll_tension_control","operating_system":"Linux","task_description":"Objective: Implement and validate a deterministic feedback controller for a six-section roll-to-roll web line that tracks changing web-tension and line-speed references under uncertain material and roller parameters.\n\nStarting State: A read-only input folder containing the task and success specifications (TASK.md, SUCCESS.md), the complete public nonlinear plant with rollout, safety, and metric code (benchmark.py — no hidden stabilizer), the required controller interface (controller_template.py), two public validation scenarios (public_cases.json), a public evaluation harness (run_public_eval.py), and a pinned runtime spec (input/runtime_env/pyproject.toml).\n\nInstructions & Rules: Copy controller_template.py to the output directory as controller.py and implement the Controller API (reset(), and act() returning six motor torques). The controller must be deterministic, must not read plant internals, must keep every torque within +/-12 N m, must never violate the plant safety limits, and must average under 5 ms per act() call. Run run_public_eval.py on both public cases and write exactly four files to the output directory: controller.py, results.json, trajectories.csv, validation_report.md. Numeric claims in the report must match the artifacts. The official evaluator reruns the controller on six hidden combinations of material stiffness, inertia, roller radius, friction, initial-state error, reference changes, seeded process noise, and transient disturbances; it does not trust self-reported performance.","software_licensing":"Free/Open Source","evaluation_criteria":null,"verification_method":"Deterministic scorer in [0, 1], no LLM judge. A runnable grader is included in the reference upload (a complete ALE task package: task_card.json, main.py, scripts/verify_submission.py).\n\nHard gates (any failure scores 0.0): exact output structure; Controller API and action shape; finite values; +/-12 N m actuator bounds; plant safety limits; deterministic replay; public-artifact consistency (the submitted results.json and trajectories.csv must replay bit-consistently from the submitted controller); mean act() runtime <= 5 ms including isolated-worker IPC overhead.\n\nScore: the evaluator reruns the submitted controller on six withheld cases. 10% of the score is public-artifact consistency; 90% compares hidden tension RMSE, velocity RMSE, peak tension error, control RMS, and control variation against the included genuine reference controller. Missing or invalid output returns 0.0; infrastructure failures raise instead of penalizing the submission.\n\nAlready validated end-to-end in the official ALE Docker lifecycle at ALE commit 0b6465b: the reference controller scores 1.0 (all gates, 6/6 hidden cases) and a deliberately invalid NaN controller scores 0.0.","software_and_version":"Python 3.12, NumPy 2.5.2, SciPy 1.18.1","specific_requirements":null,"expected_completion_time":null},"files":[{"file_id":"7f6a8c51-b2d3-433c-a2fb-dd3c87eaf5ff","filename":"r2r_input_materials.zip","category":"input_materials","content_type":"application/zip","size_bytes":9031},{"file_id":"dff38e1a-4e62-4e23-94bc-99ab35de42c7","filename":"r2r_ale_evaluator_handoff.tar.gz","category":"reference_output","content_type":"application/x-gzip","size_bytes":248135}],"deep_review":{"overall_verdict":"Strong Accept","review_text":"**Overall Verdict:** Strong Accept — The submission is complete, technically coherent, professionally challenging, objectively scored, and reachable end to end without material risk.\n\n**Completeness Check:**\n- Instruction: Complete — The task, controller API, state ordering, actuator and safety limits, determinism, runtime constraint, and exact deliverables are explicit.\n- Inputs: Complete — The complete nonlinear plant, controller template, two public scenarios, public evaluation harness, and success specification are present and consistent.\n- Toolchain: Complete — A pinned Python, NumPy, and SciPy environment is specified for CPU-only Linux using freely obtainable software.\n- Reference output: Complete — The private handoff includes a genuine controller, replay-consistent results, a 32-column trajectory with 1,300 valid rows, a matching report, six hidden cases, and complete baseline metrics.\n- Rubric: Complete — The evaluator implements documented hard gates and an explicit weighted score across tracking error, peak error, control effort, and variation.\n\n**File and Task Reachability:**\nAttachment classification: reachable. Task classification: reachable. Both `r2r_input_materials.zip` and `r2r_ale_evaluator_handoff.tar.gz` were inspected. Their byte counts and actual formats match the manifest; both containers are readable, unencrypted, modestly sized, and free of unsafe paths, links, or device entries. All members were inventoried and statically inspected, including specifications, plant and harness source, public and hidden scenario schemas, evaluator source, task metadata, licensing and provenance records, reference controller, results, report, and the complete trajectory dataset. The handoff’s internal checksums all match, and its public-input copies are byte-identical to the solver archive. No relevant content remained unreviewed. Uploaded programs were not executed under safe-review rules, so the documented lifecycle runs were not independently reproduced; static inspection nevertheless establishes a complete implementation and evaluation path. Hidden cases, baselines, and reference outputs are clearly separated from solver-visible materials, and evaluator isolation removes hidden files before candidate execution. License keys, accounts, tokens, and other credentials are not required from the contributor. Confidential, proprietary, or internal material may be evaluated as submitted evidence without requiring additional disclosure.\n\n**Special Screening:**\nNone\n\n**Recommended Changes:**\nThe submission provides unusually complete specifications, artifacts, provenance, isolation controls, and validation evidence. No changes are required.","reviewed_at":"2026-09-02T03:15:19.890498Z"},"claim":{"id":"07815cf9-8994-4b56-b430-aa92f83f841f","submission_id":"3a810413-5841-4e3a-85b9-e9593d87908e","submission_revision_id":"94550bf6-0223-4844-a0fd-e6601d26bb6b","builder":"builder-01","status":"active","claimed_at":"2026-10-15T15:17:24.081153Z","delivered_at":null,"ended_at":null,"delivery_note":null,"end_reason":null}},"revision_is_current":true}