LoCoMo

AI Agents In-Place TTT LLM LoCoMo LongMemEval Memory Benchmarks Test-Time Training

Vier Memory-Benchmarks, eine Full-Context-Obergrenze von 72.9% und ein ICLR-2026-Paper, das Memory zurück in die Modellgewichte holt.

24.04.2026 11 min

AI Agents In-Place TTT LLM LoCoMo LongMemEval Memory Benchmarks Test-Time Training

Four memory benchmarks, a full-context ceiling of 72.9%, and an ICLR 2026 paper that pulls memory back into the model weights.

24.04.2026 12 min