# Uncertainty Calibration on TriviaQA + Natural Questions ID held-out (test)

Source: [Uncertainty Calibration on TriviaQA + Natural Questions ID held-out (test)](<https://www.sota2.com/research/sota/uncertainty-calibration-on-triviaqa-plus-natural-questions-id-held-out-test>)

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

Data updated: 2026-03-09T08:46:56.367Z

Evaluation of model uncertainty calibration performance using metrics like ECE, AUROC, and Spearman correlation on the held-out test split combining TriviaQA and Natural Questions datasets.

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 | Uncertainty Calibration |
| Dataset (including any stated split) | TriviaQA + Natural Questions ID held-out (test) |
| Metric | ECE |
| Unit | PERCENT |
| Direction | Lower 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

- [ECE](<https://www.sota2.com/research/sota/uncertainty-calibration-on-triviaqa-plus-natural-questions-id-held-out-test.md?metric=ECE>): Lower is better; unit: PERCENT.

- [AUROC](<https://www.sota2.com/research/sota/uncertainty-calibration-on-triviaqa-plus-natural-questions-id-held-out-test.md?metric=AUROC>): Higher is better; unit: PERCENT.

- [Spearman Correlation](<https://www.sota2.com/research/sota/uncertainty-calibration-on-triviaqa-plus-natural-questions-id-held-out-test.md?metric=Spearman%20Correlation>): Higher is better; unit: SCALAR.

## Recorded results

| Method | ECE | Result date | Report paper |
| --- | --- | --- | --- |
| Entropy-based (ours) | 7.2 | Not recorded | 69ae53e82d603e6de3e42997 |
| Brier | 15.7 | Not recorded | 69ae53e82d603e6de3e42997 |
| Base+CoT | 34.17 | Not recorded | 69ae53e82d603e6de3e42997 |
| Base | 41.99 | Not recorded | 69ae53e82d603e6de3e42997 |

### Entropy-based (ours)

ECE: 7.2. Result ID: 69ae88ea6dbb7075b956bc41.

[Reporting paper](<https://www.sota2.com/research/paper/69ae53e82d603e6de3e42997.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69ae53e82d603e6de3e42997.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B-Instruct","Reinforcement learning":"GRPO","Reward design":"Calibrated entropy-based reward signal"}. This record does not assert independent reproduction.

### Brier

ECE: 15.7. Result ID: 69ae88ea6dbb7075b956bc40.

[Reporting paper](<https://www.sota2.com/research/paper/69ae53e82d603e6de3e42997.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a241e209131cea81436.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B-Instruct","Reinforcement learning":"GRPO","Reward design":"Brier score"}. This record does not assert independent reproduction.

### Base+CoT

ECE: 34.17. Result ID: 69ae88ea6dbb7075b956bc3f.

[Reporting paper](<https://www.sota2.com/research/paper/69ae53e82d603e6de3e42997.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e571e209131cea3f27b.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B-Instruct","Reinforcement learning":"None","Chain-of-thought reasoning":"true"}. This record does not assert independent reproduction.

### Base

ECE: 41.99. Result ID: 69ae88ea6dbb7075b956bc3e.

[Reporting paper](<https://www.sota2.com/research/paper/69ae53e82d603e6de3e42997.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B-Instruct","Reinforcement learning":"None","Chain-of-thought reasoning":"false"}. This record does not assert independent reproduction.

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