NimCritic LOOM Direct Comparison - Overall PRC-F1: leaderboard
NimCritic v1 visual manipulation-progress evaluation, not action execution. LOOM: 300 trajectories, 19,618 calibrated frames, 171 single-task and 129 multi-task; duration-weighted Spearman forward/reverse correlations, signed harmonic PRC-F1 (zero on opposite signs) and terminal completion accuracy with reference progress > 0.99. Native baseline interfaces/output definitions retained. GVL shuffles up to 30 frames, temperature 1; GPT-6 Sol extra reasoning disabled. NimCritic is a separately trained Wan2.2-TI2V-5B-initialized critic with frozen UMT5-XXL,128,000 updates and 20 Euler inference steps; Gemini 3.8 Flash decomposes subtasks. Other checkpoint revisions/efforts unreported; no aliases to untrained backbones. Same critic, whole-video direct scoring versus Gemini segmentation; held separately from main comparison. Metric: PRC-F1.
Metric: PRC-F1. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | NimCritic (Harness: Gemini 3.8 Flash subtask decomposition) | 0.86 | |
| 2 | NimCritic study NimCritic w/o Subtask Decomposition | 0.59 |
Interactive version: theaggregate.ai/benchmark?slug=nimcritic-loom-direct-comparison-overall-prc-f1 · How It Works · Data refreshed daily, snapshot 2026-10-27.