# Hallucination Detection on HaluEval Gemini outputs (test)

Source: [Hallucination Detection on HaluEval Gemini outputs (test)](<https://www.sota2.com/research/sota/hallucination-detection-on-halueval-gemini-outputs-test>)

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

Data updated: 2026-06-02T13:12:50.395Z

Evaluation of model performance on Hallucination Detection using the test split of HaluEval, specifically focusing on outputs generated by Gemini models, 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 | Hallucination Detection |
| Dataset (including any stated split) | HaluEval Gemini outputs (test) |
| Metric | AUROC |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| 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](<https://www.sota2.com/research/sota/hallucination-detection-on-halueval-gemini-outputs-test.md?metric=AUROC>): Higher is better; unit: SCALAR.

## Recorded results

| Method | AUROC | Result date | Report paper |
| --- | --- | --- | --- |
| Semantic Neg. | 0.571 | Not recorded | 6a1e641fdbd7150506f8577f |
| Sem. Sets Conf. | 0.571 | Not recorded | 6a1e641fdbd7150506f8577f |
| Non-Cont. | 0.55 | Not recorded | 6a1e641fdbd7150506f8577f |
| GPT-4o judge | 0.541 | Not recorded | 6a1e641fdbd7150506f8577f |
| Entailment | 0.54 | Not recorded | 6a1e641fdbd7150506f8577f |
| Cosine Sim. | 0.54 | Not recorded | 6a1e641fdbd7150506f8577f |
| Exact Match | 0.533 | Not recorded | 6a1e641fdbd7150506f8577f |
| Sequence Prob. | 0.53 | Not recorded | 6a1e641fdbd7150506f8577f |
| P(True) | 0.528 | Not recorded | 6a1e641fdbd7150506f8577f |
| Min Token Prob. | 0.525 | Not recorded | 6a1e641fdbd7150506f8577f |
| Sonnet judge | 0.525 | Not recorded | 6a1e641fdbd7150506f8577f |
| Mean Token Neg. | 0.521 | Not recorded | 6a1e641fdbd7150506f8577f |
| Prob. Margin | 0.52 | Not recorded | 6a1e641fdbd7150506f8577f |
| Gemini judge | 0.514 | Not recorded | 6a1e641fdbd7150506f8577f |
| Llama judge | 0.506 | Not recorded | 6a1e641fdbd7150506f8577f |

### Semantic Neg.

AUROC: 0.571. Result ID: 6a1ed6b7fbc373592c1ad302.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

Recorded evaluation configuration: {"Scorer Category":"BB"}. This record does not assert independent reproduction.

### Sem. Sets Conf.

AUROC: 0.571. Result ID: 6a1ed6b7fbc373592c1ad304.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f141e209131cea43bf3.md>)

Recorded evaluation configuration: {"Scorer Category":"BB"}. This record does not assert independent reproduction.

### Non-Cont.

AUROC: 0.55. Result ID: 6a1ed6b7fbc373592c1ad301.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70d71e209131cea4d943.md>)

Recorded evaluation configuration: {"Scorer Category":"BB"}. This record does not assert independent reproduction.

### GPT-4o judge

AUROC: 0.541. Result ID: 6a1ed6b7fbc373592c1ad30c.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71aa1e209131cea526df.md>)

[Implementation](<https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge>)

Recorded evaluation configuration: {"Scorer Category":"Judge"}. This record does not assert independent reproduction.

### Entailment

AUROC: 0.54. Result ID: 6a1ed6b7fbc373592c1ad303.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70d71e209131cea4d943.md>)

Recorded evaluation configuration: {"Scorer Category":"BB"}. This record does not assert independent reproduction.

### Cosine Sim.

AUROC: 0.54. Result ID: 6a1ed6b7fbc373592c1ad305.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70d71e209131cea4d943.md>)

Recorded evaluation configuration: {"Scorer Category":"BB"}. This record does not assert independent reproduction.

### Exact Match

AUROC: 0.533. Result ID: 6a1ed6b7fbc373592c1ad306.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70d71e209131cea4d943.md>)

Recorded evaluation configuration: {"Scorer Category":"BB"}. This record does not assert independent reproduction.

### Sequence Prob.

AUROC: 0.53. Result ID: 6a1ed6b7fbc373592c1ad309.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa694d1e209131cea21611.md>)

Recorded evaluation configuration: {"Scorer Category":"WB"}. This record does not assert independent reproduction.

### P(True)

AUROC: 0.528. Result ID: 6a1ed6b7fbc373592c1ad307.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6df71e209131cea3d034.md>)

Recorded evaluation configuration: {"Scorer Category":"WB"}. This record does not assert independent reproduction.

### Min Token Prob.

AUROC: 0.525. Result ID: 6a1ed6b7fbc373592c1ad30b.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa694d1e209131cea21611.md>)

Recorded evaluation configuration: {"Scorer Category":"WB"}. This record does not assert independent reproduction.

### Sonnet judge

AUROC: 0.525. Result ID: 6a1ed6b7fbc373592c1ad30d.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71aa1e209131cea526df.md>)

[Implementation](<https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge>)

Recorded evaluation configuration: {"Scorer Category":"Judge"}. This record does not assert independent reproduction.

### Mean Token Neg.

AUROC: 0.521. Result ID: 6a1ed6b7fbc373592c1ad308.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa694d1e209131cea21611.md>)

Recorded evaluation configuration: {"Scorer Category":"WB"}. This record does not assert independent reproduction.

### Prob. Margin

AUROC: 0.52. Result ID: 6a1ed6b7fbc373592c1ad30a.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a011e209131cea254ca.md>)

Recorded evaluation configuration: {"Scorer Category":"WB"}. This record does not assert independent reproduction.

### Gemini judge

AUROC: 0.514. Result ID: 6a1ed6b7fbc373592c1ad30f.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71aa1e209131cea526df.md>)

[Implementation](<https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge>)

Recorded evaluation configuration: {"Scorer Category":"Judge"}. This record does not assert independent reproduction.

### Llama judge

AUROC: 0.506. Result ID: 6a1ed6b7fbc373592c1ad30e.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e641fdbd7150506f8577f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71aa1e209131cea526df.md>)

[Implementation](<https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge>)

Recorded evaluation configuration: {"Scorer Category":"Judge"}. This record does not assert independent reproduction.

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