Search papers, labs, and topics across Lattice.
2
0
5
0
ContrAgent is presented, a contract-based framework for symbolic temporal supervision of LLM agents that matches state-of-the-art LLM-judge and rule-based guardrail baselines while producing deterministic, reproducible verdicts and, in the online mode, orders-of-magnitude lower per-call latency.
Standardized agent-to-agent interfaces can enable reproducible and auditable benchmarks for logical reasoning, revealing that auto-formalization agents can outperform chain-of-thought reasoning on FOLIO.
