BuilderBuilder
Benchmark family

Two-handed puzzle

Two-handed puzzle tests. Unaided runs and the decision loop keep their own prompt records, runs, and status.

Each stage, prompt version, and workflow stays in its own section, with that section's prompt receipt and status.

← All benchmarks
Two-handed puzzle public evidence
Two-handed puzzle public evidence
2Variants
7Attached runs
3Related records
0Released material sets
Unaided

Unaided

1 variant keeps its own prompt receipt and status.

Two-handed puzzle

Public evidence · prompt hash recorded

6 runs on prompt 2fd1cb6ca345

Task claw-unlock-geometry

SHA-256 2fd1cb6ca34572db43df2bb5d865f55fcb30756eddd9fa4dc203d2b9f4b74046

Exact prompt file path was not published.

Published Date unreported

No released Builder download is published for this variant.

Open this variant record

Material gaps

  • No released Builder project, harness package, or skill download.
  • Exact prompt file path was not published with this test.
  • 3 related records are listed separately and are not attached to this prompt.
  • Repeat-run reliability evidence is not established.

Attached runs

  • GPT-6 Astrajudge_pass_independently_replayed · PASS · Prompt version v1Prompt version v1SHA-256 2fd1cb6ca34572db43df2bb5d865f55fcb30756eddd9fa4dc203d2b9f4b74046
  • GPT-6 Soljudge_fail_independently_replayed · PASS · Prompt version v1Prompt version v1SHA-256 2fd1cb6ca34572db43df2bb5d865f55fcb30756eddd9fa4dc203d2b9f4b74046
  • Grok 4.6judge_fail_independently_replayed · PASS · Prompt version v1Prompt version v1SHA-256 2fd1cb6ca34572db43df2bb5d865f55fcb30756eddd9fa4dc203d2b9f4b74046
  • Grok 4.7judge_fail_independently_replayed · PASS · Prompt version v1Prompt version v1SHA-256 2fd1cb6ca34572db43df2bb5d865f55fcb30756eddd9fa4dc203d2b9f4b74046
  • MiMo V2.6 Projudge_fail_independently_replayed · pass · Prompt version v1Prompt version v1SHA-256 2fd1cb6ca34572db43df2bb5d865f55fcb30756eddd9fa4dc203d2b9f4b74046
  • Claude Opus 5.5judge_fail_independently_replayed · PASS · Prompt version v1Prompt version v1SHA-256 2fd1cb6ca34572db43df2bb5d865f55fcb30756eddd9fa4dc203d2b9f4b74046

Separate records

  • GPT-6 Astra + Jev 1.13judge_pass_independently_replayed · Prompt version v1 · Prompt hash unavailablePrompt version v1Prompt hash was not published for this record.Prompt hash was not recorded on this ledger, so it is not attached to the prompt SHA.
  • GPT-6.1 Soljudge_pass_independently_replayed · PASS · Prompt hash unavailablePrompt version was not published.Prompt hash was not published for this record.This Launch 006 record is outside the synced harness expansion, so its prompt hash is not available and it is not attached to the prompt SHA.
  • Claude Sonnet 5.5judge_fail_independently_replayed · FAIL · Prompt hash unavailablePrompt version was not published.Prompt hash was not published for this record.This Launch 006 record is outside the synced harness expansion, so its prompt hash is not available and it is not attached to the prompt SHA.
Decision loop

Decision loop

1 variant keeps its own prompt receipt and status.

Two-handed puzzle · decision loop

Decision loop · public evidence

1 run on prompt 2fd1cb6ca345

Task claw-unlock-geometry

SHA-256 2fd1cb6ca34572db43df2bb5d865f55fcb30756eddd9fa4dc203d2b9f4b74046

Exact prompt file path was not published.

Published Date unreported

No released Builder download is published for this variant.

Open this variant record

Material gaps

  • No released Builder project, harness package, or skill download.
  • Exact prompt file path was not published with this test.
  • Repeat-run reliability evidence is not established.

Attached runs

  • Jev 1.13judge_verified_decision_loop_run_independently_replayed · PASS · Prompt version v1Prompt version v1SHA-256 2fd1cb6ca34572db43df2bb5d865f55fcb30756eddd9fa4dc203d2b9f4b74046

Separate records

No separate unattached record was published for this variant.