LoRA Projection Utility Audit - Code: leaderboard
LoRA Projection v1 Table9 no-adapter PureBaseW0. Objective100question multiplechoice utilityaudit; nativepercent. No inferred standardbenchmarkname; source-unreported decoding. Poisoned/purified/retrainedadapters andASR/PPL/format diagnostics are not base model observations.
Metric: Score (%). Source: arxiv.org. 1 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen2.5-0.5B-Instruct | 48 |
Interactive version: theaggregate.ai/benchmark?slug=lora-projection-utility-audit-code · How It Works · Data refreshed daily, snapshot 2026-11-12.