Interlock: Typed Completion Graphs Push Long-Horizon Agents from Local Correctness to Task Completion
meituan-longcat · hf · 2026-10-20
Researchers from PKU, HKUST(GZ), Tsinghua and Meituan propose Interlock, addressing the completion gap where long-horizon agents produce locally correct but incomplete outcomes.
- Tasks are synthesized as typed completion graphs: grounding edges link public evidence to decisions, coupling hyperedges impose joint constraints across components, and realization edges connect decisions to all required downstream effects
- The graph compiles into a resettable workspace, a public request, and an evaluator
- Agent-based verification and self-training then improve performance on complex benchmarks
More from coding & agent
- Indie dev says AI removed the model bottleneck: solo-building a physics excavator sim — Michael_J_Black · 2026-10-20
- HSPR Recovers Full Agent Benchmark Scores From Small Subsets via Hierarchical Posterior Inference — SUDA · 2026-10-20
- AdaExploreBench Finds LLM Agents Struggle at Proactive Exploration in Evolving Environments — HKUST · 2026-10-20
- Our OAuth discovery doc and Python docs give agents opposite signup instructions — bshivarthy · 2026-10-20
- Fo launches MCP to let AI agents control and sense physical devices, 100+ devs already building — heyshrutimishra · 2026-10-20
- Open-source humanoid Asimov 1 ships docs: assembly guides, sim-to-real training, browser demo — Sentdex · 2026-10-20