Selective Memory Retention for Long-Horizon LLM Agents
About
When does retention matter for memory-augmented LLM agents? We study this with TraceRetain, a lightweight framework for bounded external memory in frozen LLM agents that scores entries by interpretable features (success, age, access frequency, redundancy, specificity, similarity, downstream utility) and evicts the lowest-scoring ones at capacity. On clean ALFWorld with gpt-5-mini, external memory robustly improves over no memory across two seeds, but differences among bounded retention policies fall within Wilson 95% CIs: clean ALFWorld at T=100 to T=200 does not naturally exhibit the memory pollution retention is designed to address. Under a controlled noisy-write stress (75% synthetic distractors), unbounded memory and FIFO-K50 degrade on Precision@5 (20.2% to 12.4% and 15.8% to 3.8%) while TraceRetain-CEM is essentially unchanged (16.9% to 16.6%) and preserves 97/100 task success. The mechanism: unbounded memory has the highest mean similarity (0.87) but lowest precision, indicating failed distractors close to the query in embedding space. Held-out in-distribution evaluation shows memory-augmented policies solving 47 to 49 of 50 tasks vs. 39/50 for no memory. Bounded retention buys memory and step efficiency on saturated clean benchmarks at no task-success cost, and only differentiates from cache heuristics when streams contain noise.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| ALFWorld task success | ALFWorld Noisy (T=100) | Task Success Count97 | 5 | |
| ALFWorld task success | ALFWorld Eval-seen (50 tasks) | Success Count (Seen Tasks)49 | 5 | |
| Task success | ALFWorld 100 tasks (train) | Success Rate97 | 5 | |
| Task success | ALFWorld (50 eval tasks) | Success Rate49 | 5 | |
| ALFWorld task success | ALFWorld Clean (T=100) | Task Success Rate97.5 | 5 | |
| ALFWorld task success | ALFWorld Clean (T=200) | Task Success Count0.96 | 5 |