feat(sob): SOB benchmark, evaluation pipeline, leaderboard

[Abhinavexist]

Initial benchmark drop. CI builds leaderboard.json from data/evaluation/ and posts a top-10 preview comment.

[github-actions]

🏆 Leaderboard preview

Built 21 models, top 10 by Overall:

Rank	Model	Overall	Val. Acc.	JSON Pass	Perfect
1	GPT-5.4	0.870	0.798	0.993	0.469
2	GLM-4.7	0.861	0.804	0.965	0.508
3	Qwen3.5-35B	0.861	0.801	0.969	0.500
4	Gemini-2.5-Flash	0.860	0.796	0.972	0.498
5	Qwen3-235B	0.857	0.786	0.978	0.463
6	Interfaze-Beta	0.855	0.795	0.967	0.480
7	Claude-Sonnet-4.6	0.854	0.779	0.979	0.442
8	GPT-4.1	0.850	0.783	0.969	0.454
9	GPT-5	0.849	0.769	0.983	0.398
10	Gemma-3-27B	0.847	0.777	0.969	0.454

Generated at 2026-04-28T08:23:42+00:00 • full JSON in workflow artifacts