# Prefill-stage hallucination risk detection on Benchmark-500 Relaxed Consensus (Pvote ≥ 0.8)

Source: [Prefill-stage hallucination risk detection on Benchmark-500 Relaxed Consensus (Pvote ≥ 0.8)](<https://www.sota2.com/research/sota/prefill-stage-hallucination-risk-detection-on-benchmark-500-relaxed-consensus-pvote-0-8>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-03-23T11:14:19.220Z

Evaluation of model performance on detecting prefill-stage hallucination risk using the Benchmark-500 Relaxed Consensus dataset, typically measured by AUROC.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Prefill-stage hallucination risk detection |
| Dataset (including any stated split) | Benchmark-500 Relaxed Consensus (Pvote ≥ 0.8) |
| Metric | AUROC (Mean) |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 4 of 4 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [AUROC (Mean)](<https://www.sota2.com/research/sota/prefill-stage-hallucination-risk-detection-on-benchmark-500-relaxed-consensus-pvote-0-8.md?metric=AUROC%20(Mean)>): Higher is better; unit: SCALAR.

- [AUROC (95% CI Lower Bound)](<https://www.sota2.com/research/sota/prefill-stage-hallucination-risk-detection-on-benchmark-500-relaxed-consensus-pvote-0-8.md?metric=AUROC%20(95%25%20CI%20Lower%20Bound)>): Higher is better; unit: SCALAR.

## Recorded results

| Method | AUROC (Mean) | Result date | Report paper |
| --- | --- | --- | --- |
| Risk-Cos | 0.6957 | Not recorded | 69c0c90e0636e5bf4afc3c63 |
| Risk-Margin | 0.673 | Not recorded | 69c0c90e0636e5bf4afc3c63 |
| Risk-Entropy | 0.502 | Not recorded | 69c0c90e0636e5bf4afc3c63 |
| Risk-Loss | 0.326 | Not recorded | 69c0c90e0636e5bf4afc3c63 |

### Risk-Cos

AUROC (Mean): 0.6957. Result ID: 69c11f5447f34ce924a9787c.

[Reporting paper](<https://www.sota2.com/research/paper/69c0c90e0636e5bf4afc3c63.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c0c90e0636e5bf4afc3c63.md>)

Recorded evaluation configuration: {"generation\_mode":"prefill only"}. This record does not assert independent reproduction.

### Risk-Margin

AUROC (Mean): 0.673. Result ID: 69c11f5447f34ce924a9787d.

[Reporting paper](<https://www.sota2.com/research/paper/69c0c90e0636e5bf4afc3c63.md>)

Recorded evaluation configuration: {"generation\_mode":"prefill only"}. This record does not assert independent reproduction.

### Risk-Entropy

AUROC (Mean): 0.502. Result ID: 69c11f5447f34ce924a9787e.

[Reporting paper](<https://www.sota2.com/research/paper/69c0c90e0636e5bf4afc3c63.md>)

Recorded evaluation configuration: {"generation\_mode":"prefill only"}. This record does not assert independent reproduction.

### Risk-Loss

AUROC (Mean): 0.326. Result ID: 69c11f5447f34ce924a9787f.

[Reporting paper](<https://www.sota2.com/research/paper/69c0c90e0636e5bf4afc3c63.md>)

Recorded evaluation configuration: {"generation\_mode":"prefill only"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.