MAPE Study - LongMemEval User: leaderboard
MAPE evaluates frozen GPT-4.1 Mini and Qwen3-30B-A3B-Instruct-2507 with different memory programs. LoCoMo uses 1,307 evaluation questions and LongMemEval-S uses 400, with question-weighted accuracy. The evaluation partition also selects the final memory program. The RQS transplant experiment is a separate rerun. This board reports lLM-judged correctness (%). Higher is better.
Metric: LLM-judged correctness (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 1 model tracked. Updated 14 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Qwen3-30B-A3B-Instruct-2507 | Evaluation run: MAPE LightMem | 96.43 |
Interactive version: theaggregate.ai/benchmark?slug=mape-study-longmemeval-user · How It Works · Data refreshed daily, snapshot 2026-11-16.