Google's ProceduralGraphs wires tools, skills and memory into evolvable graphs, lifting agent survival 6% to 34%
新智元 · wechat · 2026-09-21
Researchers from Google, Georgia Tech and Peking University propose ProceduralGraphs (PG), representing tool calls, skills, reasoning and task states as a directed graph of "process–relation–process" triples whose edges carry conditions, guidance and pitfalls. Paper: arxiv.org/pdf/2609.09153
Design: at runtime the agent locates its current node, reads the local subgraph within two outgoing hops, and generates contextual guidance; offline, a revision model proposes graph edits validated against held-out tasks before adoption — no weight updates needed. Notably, a hand-written expert graph hurt MultiChallenge success (87.5%→58.93%) while evolved graphs reached 92.86%.
Results: BFCL v3 accuracy 67% vs 58% best baseline; in EnterpriseArena's 132-month business simulation, Gemini 3.1 Pro survival rose 6%→34% and Claude Sonnet 4.6 44%→58%. Local-subgraph guidance beat full-graph guidance on ALFWorld (54.48%→81.53%) with 70.9% fewer tokens. Limitations: small gains on QA tasks and extra offline compute cost.
Related event: Google's Procedural Graphs boost LLM agent long-task performance(2 posts)→
More from coding & agent
- AutoClip: AI video clipping tool hits ~8K stars on GitHub — zhouxiaoka · 2026-09-21
- Aggregate WER hid the real problem: half of bank code reads were wrong in production voice agent — Adventurous_Whole973 · 2026-09-21
- Cheapest ways to run agent loops in 2026: pay-per-token beats free tiers and subs — my-coffee-where · 2026-09-21
- Fine-tune your own model for free: Claude Code writes the code, Kaggle T4s do the training — ojasvi_yadav · 2026-09-21
- Tencent open-sources T-Mem, a memory architecture that anticipates instead of archiving — blaizedsouza · 2026-09-21
- Ingesting 100k papers into an LLM wiki: user seeks editorial policy for AI knowledge synthesis — gintrux · 2026-09-21