Metacognitive Control - HotpotQA: leaderboard
Metacognitive Control compares decoding and tool-use strategies for LLaDA model variants on function calling and multi-hop questions. Cover exact match (%).
Metric: Cover exact match (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Metacognitive-controller LLaDA-8B | 46.6 | |
| 2 | Metacognitive-paper LLaDA-8B SFT | 45.7 | |
| 3 | Metacognitive-paper LLaDA-8B SFT Fast-dLLM | 43.2 | |
| 4 | DLLM-Searcher | 37.1 | |
| 5 | LLaDA2.0-mini | 36.3 | |
| 6 | LLaDA-8B-Tools | 16.5 |
Interactive version: theaggregate.ai/benchmark?slug=metacognitive-control-hotpotqa · How It Works · Data refreshed daily, snapshot 2026-10-29.