Handed a 300-Email Procurement Inbox, the Agent Must Decide What's True and Buy
alifcoder · x · 2026-08-26
The procurement test in CommerceAgentBench hands the agent an inbox of roughly 300 emails, buried with supplier quotes, revised quantities, delivery terms, payment details, withdrawn quotations, and similar-looking supplier identities. The task is not to "summarize these emails" but to figure out what is actually true, narrow down valid options, and support a final buying decision based on rules established by a human.
More from Research
- LAWA: Efficient Future Imagination for World Models via Latent Actions — Xiang Li · 2026-08-26
- FaSHIoNPOp: Fast Spectrally-Shaped Disorder Generation on the Sphere — ssh4net · 2026-08-26
- Inflatable Robotic Skin with Internal ToF Sensing Debuts at IROS 2026 — nptacek · 2026-08-26
- Developer builds local agent "Secret Agent Bubbles": Qwen works by day, self-trains by RL at night — cephaloform · 2026-08-26
- Why mathematicians lack motivation to verify AI proofs — voooooogel · 2026-08-26
- Skepticism Around LLM-as-a-Judge: Biases and Overfitting Risks — Innowise_ · 2026-08-26