MIT Releases FinanceGym, Largest Finance Agent Dataset; Even Top Models Fail

MIT CSAIL released FinanceGym, the largest finance agent dataset with over 59,000 samples spanning 37 roles and 40 apps—10 times larger than prior benchmarks. Testing showed even the strongest models, including Kimi and Qwen, failed to pass.

2026-08-25 ~ 2026-08-25 · 4 related posts

• MIT Releases FinanceGym Dataset; Kimi and Qwen Fail Tests — MIT_CSAIL · 2026-08-25

3 near-duplicate retellings: MIT_CSAIL · qinzytech · qinzytech