# Hallucination Detection on TriviaQA n=1000 (test)

Source: [Hallucination Detection on TriviaQA n=1000 (test)](<https://www.sota2.com/research/sota/hallucination-detection-on-triviaqa-n-1000-test>)

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

Data updated: 2026-05-07T10:03:59.924Z

Evaluation of model performance on Hallucination Detection using the specific n=1000 subset of the TriviaQA test split.

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) | TriviaQA n=1000 (test) |
| Metric | AUROC |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 24 of 24 |
| 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-triviaqa-n-1000-test.md?metric=AUROC>): Higher is better; unit: SCALAR.

- [Delta](<https://www.sota2.com/research/sota/hallucination-detection-on-triviaqa-n-1000-test.md?metric=Delta>): Higher is better; unit: SCALAR.

- [Accuracy](<https://www.sota2.com/research/sota/hallucination-detection-on-triviaqa-n-1000-test.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | AUROC | Result date | Report paper |
| --- | --- | --- | --- |
| phi\_first | 0.794 | Not recorded | 69fc1cc45497b3b670c88788 |
| Sem. AU | 0.778 | Not recorded | 69fc1cc45497b3b670c88788 |
| Verbalized | 0.774 | Not recorded | 69fc1cc45497b3b670c88788 |
| phi\_first | 0.772 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-3w | 0.77 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-full | 0.769 | Not recorded | 69fc1cc45497b3b670c88788 |
| phi\_first | 0.764 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-full | 0.758 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-1w | 0.752 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-3w | 0.751 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-full | 0.748 | Not recorded | 69fc1cc45497b3b670c88788 |
| Sem. AU | 0.748 | Not recorded | 69fc1cc45497b3b670c88788 |
| Sem. AU | 0.741 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-3w | 0.741 | Not recorded | 69fc1cc45497b3b670c88788 |
| phi\_first | 0.727 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-1w | 0.725 | Not recorded | 69fc1cc45497b3b670c88788 |
| Sem. AU | 0.724 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-full | 0.718 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-1w | 0.703 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-3w | 0.701 | Not recorded | 69fc1cc45497b3b670c88788 |
| Verbalized | 0.696 | Not recorded | 69fc1cc45497b3b670c88788 |
| Verbalized | 0.695 | Not recorded | 69fc1cc45497b3b670c88788 |
| AU-1w | 0.632 | Not recorded | 69fc1cc45497b3b670c88788 |
| Verbalized | 0.614 | Not recorded | 69fc1cc45497b3b670c88788 |

### phi\_first

AUROC: 0.794. Result ID: 69fc637347f34ce924dfa13c.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Inference Cost":"1 decode"}. This record does not assert independent reproduction.

### Sem. AU

AUROC: 0.778. Result ID: 69fc637347f34ce924dfa13b.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

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

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### Verbalized

AUROC: 0.774. Result ID: 69fc637347f34ce924dfa143.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70f81e209131cea4e72e.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Inference Cost":"1 decode"}. This record does not assert independent reproduction.

### phi\_first

AUROC: 0.772. Result ID: 69fc637347f34ce924dfa148.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Inference Cost":"1 decode"}. This record does not assert independent reproduction.

### AU-3w

AUROC: 0.77. Result ID: 69fc637347f34ce924dfa139.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### AU-full

AUROC: 0.769. Result ID: 69fc637347f34ce924dfa13a.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### phi\_first

AUROC: 0.764. Result ID: 69fc637347f34ce924dfa14e.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"TriviaQA mean","Inference Cost":"1 decode"}. This record does not assert independent reproduction.

### AU-full

AUROC: 0.758. Result ID: 69fc637347f34ce924dfa146.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### AU-1w

AUROC: 0.752. Result ID: 69fc637347f34ce924dfa138.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### AU-3w

AUROC: 0.751. Result ID: 69fc637347f34ce924dfa145.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### AU-full

AUROC: 0.748. Result ID: 69fc637347f34ce924dfa14c.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"TriviaQA mean","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### Sem. AU

AUROC: 0.748. Result ID: 69fc637347f34ce924dfa14d.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

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

Recorded evaluation configuration: {"Model":"TriviaQA mean","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### Sem. AU

AUROC: 0.741. Result ID: 69fc637347f34ce924dfa147.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

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

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### AU-3w

AUROC: 0.741. Result ID: 69fc637347f34ce924dfa14b.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"TriviaQA mean","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### phi\_first

AUROC: 0.727. Result ID: 69fc637347f34ce924dfa142.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Mistral-7B-v0.3","Inference Cost":"1 decode"}. This record does not assert independent reproduction.

### AU-1w

AUROC: 0.725. Result ID: 69fc637347f34ce924dfa144.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### Sem. AU

AUROC: 0.724. Result ID: 69fc637347f34ce924dfa141.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

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

Recorded evaluation configuration: {"Model":"Mistral-7B-v0.3","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### AU-full

AUROC: 0.718. Result ID: 69fc637347f34ce924dfa140.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Mistral-7B-v0.3","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### AU-1w

AUROC: 0.703. Result ID: 69fc637347f34ce924dfa14a.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"TriviaQA mean","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### AU-3w

AUROC: 0.701. Result ID: 69fc637347f34ce924dfa13f.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Mistral-7B-v0.3","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### Verbalized

AUROC: 0.696. Result ID: 69fc637347f34ce924dfa13d.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70f81e209131cea4e72e.md>)

Recorded evaluation configuration: {"Model":"Mistral-7B-v0.3","Inference Cost":"1 decode"}. This record does not assert independent reproduction.

### Verbalized

AUROC: 0.695. Result ID: 69fc637347f34ce924dfa149.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70f81e209131cea4e72e.md>)

Recorded evaluation configuration: {"Model":"TriviaQA mean","Inference Cost":"1 decode"}. This record does not assert independent reproduction.

### AU-1w

AUROC: 0.632. Result ID: 69fc637347f34ce924dfa13e.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

Recorded evaluation configuration: {"Model":"Mistral-7B-v0.3","Inference Cost":"1+10 decodes (sampling)"}. This record does not assert independent reproduction.

### Verbalized

AUROC: 0.614. Result ID: 69fc637347f34ce924dfa137.

[Reporting paper](<https://www.sota2.com/research/paper/69fc1cc45497b3b670c88788.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70f81e209131cea4e72e.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Inference Cost":"1 decode"}. This record does not assert independent reproduction.

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