AR to diffusion conversion

After grading, a reviewer model reads the submission's files and the agent's trajectory and checks each of the task's rules. A failed check makes the score 0. Step numbers link into the trajectory.

LLM adversarial review is not proof of arbitrary program behavior. Host storage and task-specific deterministic checks remain required.

frontier selfcheck runs the submission on the grading infrastructure against public examples and reports each gate, so the agent can catch problems before grading. Each run gets four. This agent used 4 of 4.

1. 1
   passed3h 13m into the runtook 4m 30s

   startup 92.9 s of 300 s; 1/1 gates passed; The suite's 32 concurrent requests per copy, up to 3600 s each, were not sent; one request at its settings was.

   5 gates

   • ✓review prerequisites
   • ✓startup · GET /health answered 200 after 92.9 s
   • ✓single-stream speed · 8 of 8 measured prompts, sent one at a time after an untimed warm-up (max_tokens 1024, temperature 1), each reply's reasoning and answer tokens counted with the pinned openbmb/MiniCPM5-2B tokenizer over end-to-end request time (sandbox clock), a miss re-sent once; slowest 590 tokens/s (982 tokens in 1.66 s); per prompt: 590, 596, 625, 633, 685, 612, 597, 604. Public prompts: 7 GPQA and 2 hard LiveCodeBench; grading draws 4 GPQA, 3 level-5 MATH-500 (held out) and 2 hard LiveCodeBench, so its replies differ.
   • ✓GPQA request as graded (self-check cap 291 s; grading allows 3600 s) · {"finish_reason": "stop", "answer": null, "note": "no answer letter found (ANSWER: X on its own line): grading scores it wrong", "content_characters": 8830, "reasoning_characters": 536, "content_end": "rix }{\n{bmatrix{bmatrix } }{bmatrix{bmatrix =\n{bmatrix\n\n ]{bmatrix -1{b{bmatrix ...[clipped]
   • ✓healthy after requests
2. 2
   passed3h 23m into the runtook 6m 37s

   startup 83.8 s of 300 s; 1/1 gates passed; The suite's 32 concurrent requests per copy, up to 3600 s each, were not sent; one request at its settings was.

   5 gates

   • ✓review prerequisites
   • ✓startup · GET /health answered 200 after 83.8 s
   • ✓single-stream speed · 8 of 8 measured prompts, sent one at a time after an untimed warm-up (max_tokens 1024, temperature 1), each reply's reasoning and answer tokens counted with the pinned openbmb/MiniCPM5-2B tokenizer over end-to-end request time (sandbox clock), a miss re-sent once; slowest 499 tokens/s (962 tokens in 1.93 s); per prompt: 520, 531, 499, 526, 521, 518, 525, 519. Public prompts: 7 GPQA and 2 hard LiveCodeBench; grading draws 4 GPQA, 3 level-5 MATH-500 (held out) and 2 hard LiveCodeBench, so its replies differ.
   • ✓GPQA request as graded (self-check cap 291 s; grading allows 3600 s) · {"finish_reason": "length", "note": "cut off at max_tokens: grading scores such a reply zero", "content_characters": 0, "reasoning_characters": 0, "content_end": ""}
   • ✓healthy after requests
3. 3
   passed3h 33m into the runtook 4m 58s

   startup 100.2 s of 300 s; 1/1 gates passed; The suite's 32 concurrent requests per copy, up to 3600 s each, were not sent; one request at its settings was.

   5 gates

   • ✓review prerequisites
   • ✓startup · GET /health answered 200 after 100.2 s
   • ✓single-stream speed · 8 of 8 measured prompts, sent one at a time after an untimed warm-up (max_tokens 1024, temperature 1), each reply's reasoning and answer tokens counted with the pinned openbmb/MiniCPM5-2B tokenizer over end-to-end request time (sandbox clock), a miss re-sent once; slowest 497 tokens/s (983 tokens in 1.98 s); per prompt: 497, 512, 510, 523, 512, 500, 513, 509. Public prompts: 7 GPQA and 2 hard LiveCodeBench; grading draws 4 GPQA, 3 level-5 MATH-500 (held out) and 2 hard LiveCodeBench, so its replies differ.
   • ✓GPQA request as graded (self-check cap 291 s; grading allows 3600 s) · {"finish_reason": "length", "note": "cut off at max_tokens: grading scores such a reply zero", "content_characters": 0, "reasoning_characters": 0, "content_end": ""}
   • ✓healthy after requests
4. 4
   passed3h 40m into the runtook 4m 34s

   startup 96.5 s of 300 s; 1/1 gates passed; The suite's 32 concurrent requests per copy, up to 3600 s each, were not sent; one request at its settings was.

   5 gates

   • ✓review prerequisites
   • ✓startup · GET /health answered 200 after 96.5 s
   • ✓single-stream speed · 8 of 8 measured prompts, sent one at a time after an untimed warm-up (max_tokens 1024, temperature 1), each reply's reasoning and answer tokens counted with the pinned openbmb/MiniCPM5-2B tokenizer over end-to-end request time (sandbox clock), a miss re-sent once; slowest 504 tokens/s (820 tokens in 1.63 s); per prompt: 516, 507, 506, 512, 525, 513, 504, 516. Public prompts: 7 GPQA and 2 hard LiveCodeBench; grading draws 4 GPQA, 3 level-5 MATH-500 (held out) and 2 hard LiveCodeBench, so its replies differ.
   • ✓GPQA request as graded (self-check cap 291 s; grading allows 3600 s) · {"finish_reason": "stop", "answer": null, "note": "no answer letter found (ANSWER: X on its own line): grading scores it wrong", "content_characters": 12028, "reasoning_characters": 1966, "content_end": " -\\sigma$.\n = $[ ] $ ] ].\n\nThen }\n ^text{}}}$.\n]} = ...[clipped]
   • ✓healthy after requests

The agent's last message before its session ended.