Add Aletheia HumanEval / HumanEval+ and MBPP / MBPP+ official evaluate scores

[shaneraphel]

Summary

• Add Aletheia to results.json.
• link is the public page: https://github.com/shaneraphel/aletheia
• HumanEval completions for an independent evalplus.evaluate --dataset humaneval run: https://github.com/shaneraphel/aletheia/blob/main/humaneval/samples.jsonl
• MBPP completions for an independent evalplus.evaluate --dataset mbpp run: https://gist.github.com/shaneraphel/b0fbe4cde5b83185da043c9e0ac00fd0
• open-data is NONE.

Test plan

• pip install evalplus==0.3.1 && evalplus.evaluate --dataset humaneval --samples humaneval/samples.jsonl
• pip install evalplus==0.3.1 && evalplus.evaluate --dataset mbpp --samples samples.jsonl
• Confirm results.json still parses and the leaderboard page renders

[shaneraphel]

Official evalplus.evaluate 0.3.1 reprint on the same MBPP+ v0.2.0 pin, parallel=2: MBPP 100.0 / MBPP+ 86.8 (378/378 base, 328/378 plus). Entry size remains 0, open-data NONE, prompted true.

[shaneraphel]

Official evalplus.evaluate 0.3.1 reprint on the same MBPP+ v0.2.0 pin, parallel=2: MBPP 100.0 / MBPP+ 87.0 (378/378 base, 329/378 plus). Entry size remains 0, open-data NONE, prompted true.

[shaneraphel]

Public page: https://github.com/shaneraphel/aletheia

Completions (samples.jsonl, 378 rows): https://gist.github.com/shaneraphel/b0fbe4cde5b83185da043c9e0ac00fd0

SHA-256: ba41a115da629ac4bf55213d4e29e4545994e9a014bbee50fc1f4c384dd9f3b1

[shaneraphel]

HumanEval / HumanEval+ numbers are now filled from the same official evalplus.evaluate 0.3.1 path as the MBPP row.

• HumanEval pass@1 98.2 (161/164)
• HumanEval+ pass@1 93.9 (154/164)
• samples: https://github.com/shaneraphel/aletheia/blob/main/humaneval/samples.jsonl
• sha256 2509956cdf8cf7b6b8d8092a96b6df07a8fad73e12c3e44fa3ecef447db4ff7c
• release: https://github.com/shaneraphel/aletheia/releases/tag/humaneval-evalplus-20260911

MBPP / MBPP+ is unchanged: 100.0 / 87.0.