TSUMEGO / BENCHv0.1 · 15 runs

Generated problems

Benchmark runs

Select a run and problem to inspect the generated position, solution tree, and recorded checks.

Anthropic / Claude CLIreviewed

claude-fable-5

Aug 8, 2026, 3:25 PM · low · 3h 41m 49s

10 structural10 original0 to review2 human score42 / 50queries used20 / 21 / 1clear / flagged / error
ROOT · SETUPStarting position

20–30 kyu target.

GENERATED POSITION
RIGHT NO RIGHT
SOLUTION TREE10 nodes across 2 forks
Leads to RIGHT No RIGHT
SETUP17stones
DEPTH3moves max
RIGHT2accepted lines
TARGET20–30 kyu
AUTOMATED CHECKS
  • pass

    The SGF passed the automated structural and legality checks.

  • pass

    Local and remote duplicate gates passed.

  • pass

    All 10 expected SGF files are present.

  • pass

    The output directory contains only the 10 expected files.

  • pass

    The model-visible input snapshot is unchanged.

  • pass

    The set includes 6 Black-to-play and 4 White-to-play problems.

  • needs human review

    The 10 per-file difficulty targets span the benchmark range; a human reviewer must verify the actual difficulty.

  • pass

    All 10 exact final output hashes received a clear originality result.

DETAILS
Target difficulty
20–30 kyu
Human review
1 review
Closest reference
#10330 · 18%
Remote top match
71% · #39541 ↗
Human review choices
AdamReviewed
Valid
Yes
Realistic
Yes
Duplicate
No
Well pathed
No
Difficulty
Not set
Quality
Not rated
SGF SOURCEproblem-01.sgf
Reveal SGF source
(;SZ[19]FF[4]GM[1]CA[UTF-8]AP[tsumego-bench]AB[ar][cr][cs][dq][dr][er][es]AW[aq][bq][cq][cp][dp][ep][eq][fq][fr][fs]
(;W[br]
(;B[as];W[bs]C[RIGHT])
(;B[bs];W[as]C[RIGHT]))
(;W[bs];B[br])
(;W[as];B[br]))