X-Scientist harness adds +7.4 points over Codex on DrugDiscoveryBench with the same model
BoWang87 · x · 2026-10-22
Evaluated without domain-specific tools on two public benchmarks, Xaira's X-Scientist harness consistently beat comparison harnesses at equal model and reasoning effort. On DrugDiscoveryBench, with Astra it scored 72.8 (+7.4 points over Codex, +4.1 over mini-SWE-agent); with Sonnet it beat Claude Code by 13.4 points. On Terminal-Bench-Science's life sciences subset, it led Codex by 7.1 points at max effort. Cross-model, X-Scientist+Sonnet hit 58.1%, 4.4 points above Claude Code+Opus. Evaluations used best-of-three generation and verification — system-level comparisons.
More from Research
- We need an L0.5 for continual learning: agent state stacks up in 4 layers — willcb · 2026-10-22
- JOLT trains the teacher too, with penalties when its predictions diverge from the student — perplexity_ai · 2026-10-22
- Perplexity's JOLT matches GRPO with 6.5x fewer training tokens, JOLT+ wins on all five benchmarks — perplexity_ai · 2026-10-22
- Parallel Decoding Distillation Adopted by Alibaba, NVIDIA, Amazon Weeks After Release — ArashVahdat · 2026-10-22
- RFD4-Proteina hits 3.5% DNA-binder pass rate vs 0.8% for BoltzGen — AllThingsApx · 2026-10-22
- RFD4-Proteina passes computational checks on all 77 small-molecule targets — AllThingsApx · 2026-10-22