Surge AI details DAYJOB: Tax — RL environments testing agent judgment on real tax traps
echen · x · 2026-10-28
Surge AI's blog details DAYJOB: Tax, the third in its benchmark family after Healthcare and Finance, evaluating agents on tax prep, compliance and advisory. Its flagship case: an agent correctly concludes either closing date qualifies for the $250K capital-gains exclusion but must proactively flag that the two-of-five-years residency window expires around August 2027. Opus caught the deadline 5/5 times; Astra never did. The release ships new frontier data and off-the-shelf RL environments.
Related event: Surge AI Releases DAYJOB: Tax Benchmark for Real-World AI Tax Agents(3 posts)→
More from Research
- Open-source Deepthink claims agent-swarm test-time scaling lifts small-model scores from 14% to 65% — Ryoiki-Tokuiten · 2026-10-28
- All Seven Millennium Prize Problems See Breakthrough Claims, Several Driven by AI Labs — haider1 · 2026-10-28
- Fine-tuning VLAs on TAMP data hits distribution mismatch; DATAFARM gets JAX port — tomssilver · 2026-10-28
- CMU L3 Lab kicks off CUA Week with multiple realtime agent research releases — wellecks · 2026-10-28
- Math diatoms: one complex map generates a page of fractal 'microbes' — pickover · 2026-10-28
- 325K Experiments Show Personal AI Agents Steer Richer Users to Pricier Options — niloofar_mire · 2026-10-28