Search papers, labs, and topics across Lattice.
APEX-EM, a non-parametric online learning framework, was developed to equip LLM-based autonomous agents with persistent procedural memory by accumulating, retrieving, and reusing structured procedural plans without updating model weights. It employs a structured experience representation, a Plan-Retrieve-Generate-Iterate-Ingest (PRGII) workflow with Task Verifiers, and a dual-outcome Experience Memory with hybrid retrieval. Experiments on BigCodeBench, KGQAGen-10k, and Humanity's Last Exam show APEX-EM significantly improves performance, achieving up to a +48.3pp accuracy gain on KGQAGen-10k and surpassing oracle-retrieval upper bounds.
LLMs can boost their task-solving accuracy by nearly 50% simply by remembering and re-using past procedural plans, even across tasks with no lexical overlap.
LLM-based autonomous agents lack persistent procedural memory: they re-derive solutions from scratch even when structurally identical tasks have been solved before. We present \textbf{APEX-EM}, a non-parametric online learning framework that accumulates, retrieves, and reuses structured procedural plans without modifying model weights. APEX-EM introduces: (1) a \emph{structured experience representation} encoding the full procedural-episodic trace of each execution -- planning steps, artifacts, iteration history with error analysis, and quality scores; (2) a \emph{Plan-Retrieve-Generate-Iterate-Ingest} (PRGII) workflow with Task Verifiers providing multi-dimensional reward signals; and (3) a \emph{dual-outcome Experience Memory} with hybrid retrieval combining semantic search, structural signature matching, and plan DAG traversal -- enabling cross-domain transfer between tasks sharing no lexical overlap but analogous operational structure. Successful experiences serve as positive in-context examples; failures as negative examples with structured error annotations. We evaluate on BigCodeBench~\cite{zhuo2025bigcodebench}, KGQAGen-10k~\cite{zhang2025kgqagen}, and Humanity's Last Exam~\cite{phan2025hle} using Claude Sonnet 4.5 and Opus 4.5. On KGQAGen-10k, APEX-EM achieves 89.6\% accuracy versus 41.3\% without memory (+48.3pp), surpassing the oracle-retrieval upper bound (84.9\%). On BigCodeBench, it reaches 83.3\% SR from a 53.9\% baseline (+29.4pp), exceeding MemRL's~\cite{memrl2025} +11.0pp gain under comparable frozen-backbone conditions (noting backbone differences controlled for in our analysis). On HLE, entity graph retrieval reaches 48.0\% from 25.2\% (+22.8pp). Ablations show component value is task-dependent: rich judge feedback is negligible for code generation but critical for structured queries (+10.3pp), while binary-signal iteration partially compensates for weaker feedback.
