Survey: Unified Framework for Long-Horizon Agents from RUC, PKU, THU
jiqizhixin · x · 2026-08-06
Researchers from Renmin University, Peking University, and Tsinghua have released a new survey titled Towards Long-Horizon Agents.
The paper defines "long-horizon agency" as AI systems capable of extended planning, real-world interaction, learning from errors, and dynamic strategy adjustment. The authors frame this evolution as the co-evolution of an external harness and internal optimization.
A key contribution of the work is establishing the first shared taxonomy and benchmark landscape for the field, organizing scattered prior research across six perspectives from foundation to frontier.
More from coding & agent
- Perplexity's hint-guided self-distillation cuts its agent's tool-call failures by 21.2% — perplexity_ai · 2026-09-23
- Tomo ships WebMCP Registry: open index of agent-ready tools across 100K sites — Jackyhuang · 2026-09-23
- LiteParse hits 2.8ms per PDF page, 25% faster in v2.14.6, claims fastest open-source parser — llama_index · 2026-09-23
- UCLA Releases ACLArena: A Framework for Agent Continual Learning in Multi-Stage Post-Training — UCLA-SCAI · 2026-09-23
- The Hidden Cost of Using Weaker Models for Decisions: You Never Benchmark, So You Never Notice the Lost Alpha — generativist · 2026-09-23
- Coinbase opens stock trading to AI agents, with x402 micropayments for live market data — MurrLincoln · 2026-09-23