Harvey's context graphs lift legal agent scores up to 30 points and cut costs 48%
omarsar0 · x · 2026-10-16
- Harvey reports building context graphs to help agents handle long-horizon legal tasks. Agents constructed graphs from underlying documents across 11 Legal Agent Bench categories (M&A diligence, contract negotiation, etc.) without access to eval tasks or grading rubrics.
- Results: Firm Knowledge tasks 65%→80%, Precedent lists 67%→80%, Matter counting 53%→83%.
- Efficiency also improved: agents with context graph access saw a 48% cost reduction.
- dair-ai's Elvis notes structured data works well for LLM agents in complex codebases and sees broader applications; they're applying the idea to research tasks.
More from coding & agent
- Codex drives a physical CAN-bus motor, eyeing a fully Codex-built robot — OpenAIDevs · 2026-10-16
- smallwords: MIT-licensed Python lib syncing controlled vocab across GBNF, JSON Schema, validators — Empty-Reflection4280 · 2026-10-16
- Fleet dev's own lesson: 200 prod-derived eval cases still meant always playing catch-up — BraceSproul · 2026-10-16
- Box AI Extract plus OpenAI Decisions API automates document routing — craigsdennis · 2026-10-16
- OpenUI 1.0 Launches Open Source: Real-Time Generative UI Where the Question Becomes the Interface — LinusEkenstam · 2026-10-16
- The vanishing engineering signal in AI hackathons — rseroter · 2026-10-16