# Long-form generation factuality and uncertainty estimation on LongFact (test)

Source: [Long-form generation factuality and uncertainty estimation on LongFact (test)](<https://www.sota2.com/research/sota/long-form-generation-factuality-and-uncertainty-estimation-on-longfact-test>)

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

Data updated: 2026-02-26T05:45:08.620Z

Evaluation of model performance on the test split of the LongFact dataset, focusing on assessing the factuality and uncertainty estimation capabilities of long-form generated text.

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 | Long-form generation factuality and uncertainty estimation |
| Dataset (including any stated split) | LongFact (test) |
| Metric | Factuality Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 14 of 14 |
| 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

- [Factuality Score](<https://www.sota2.com/research/sota/long-form-generation-factuality-and-uncertainty-estimation-on-longfact-test.md?metric=Factuality%20Score>): Higher is better; unit: PERCENT.

- [Uncertainty Accuracy](<https://www.sota2.com/research/sota/long-form-generation-factuality-and-uncertainty-estimation-on-longfact-test.md?metric=Uncertainty%20Accuracy>): Higher is better; unit: PERCENT.

- [Inconsistency Rate](<https://www.sota2.com/research/sota/long-form-generation-factuality-and-uncertainty-estimation-on-longfact-test.md?metric=Inconsistency%20Rate>): Lower is better; unit: PERCENT.

## Recorded results

| Method | Factuality Score | Result date | Report paper |
| --- | --- | --- | --- |
| LOGU-DPO | 91.5 | Not recorded | 68fa79201e209131cea7c409 |
| LOGU-DPO | 91.3 | Not recorded | 68fa79201e209131cea7c409 |
| Unc-Zero | 89.3 | Not recorded | 68fa79201e209131cea7c409 |
| LOGU-SFT | 88.6 | Not recorded | 68fa79201e209131cea7c409 |
| Unc-Zero | 88.5 | Not recorded | 68fa79201e209131cea7c409 |
| Unc-Few | 88.3 | Not recorded | 68fa79201e209131cea7c409 |
| LOGU-SFT | 87.5 | Not recorded | 68fa79201e209131cea7c409 |
| Pair-Few | 86.7 | Not recorded | 68fa79201e209131cea7c409 |
| Orig. | 86.2 | Not recorded | 68fa79201e209131cea7c409 |
| Self-Refine | 86.2 | Not recorded | 68fa79201e209131cea7c409 |
| Pair-Few | 86.1 | Not recorded | 68fa79201e209131cea7c409 |
| Unc-Few | 85.6 | Not recorded | 68fa79201e209131cea7c409 |
| Orig. | 85.5 | Not recorded | 68fa79201e209131cea7c409 |
| Self-Refine | 85 | Not recorded | 68fa79201e209131cea7c409 |

### LOGU-DPO

Factuality Score: 91.5. Result ID: 6995c80f3badfaffe6cbbfbe.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

[Implementation](<https://github.com/rhyang2021/LoGU>)

Recorded evaluation configuration: {"Backbone":"Llama3-8B-Instruct","Method Category":"Training-Based"}. This record does not assert independent reproduction.

### LOGU-DPO

Factuality Score: 91.3. Result ID: 6995c80f3badfaffe6cbbfb7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

[Implementation](<https://github.com/rhyang2021/LoGU>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-Instruct","Method Category":"Training-Based"}. This record does not assert independent reproduction.

### Unc-Zero

Factuality Score: 89.3. Result ID: 6995c80f3badfaffe6cbbfb9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

Recorded evaluation configuration: {"Backbone":"Llama3-8B-Instruct","Method Category":"Prompt-Based"}. This record does not assert independent reproduction.

### LOGU-SFT

Factuality Score: 88.6. Result ID: 6995c80f3badfaffe6cbbfb6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

[Implementation](<https://github.com/rhyang2021/LoGU>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-Instruct","Method Category":"Training-Based"}. This record does not assert independent reproduction.

### Unc-Zero

Factuality Score: 88.5. Result ID: 6995c80f3badfaffe6cbbfb2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-Instruct","Method Category":"Prompt-Based"}. This record does not assert independent reproduction.

### Unc-Few

Factuality Score: 88.3. Result ID: 6995c80f3badfaffe6cbbfba.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

Recorded evaluation configuration: {"Backbone":"Llama3-8B-Instruct","Method Category":"Prompt-Based"}. This record does not assert independent reproduction.

### LOGU-SFT

Factuality Score: 87.5. Result ID: 6995c80f3badfaffe6cbbfbd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

[Implementation](<https://github.com/rhyang2021/LoGU>)

Recorded evaluation configuration: {"Backbone":"Llama3-8B-Instruct","Method Category":"Training-Based"}. This record does not assert independent reproduction.

### Pair-Few

Factuality Score: 86.7. Result ID: 6995c80f3badfaffe6cbbfbb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

Recorded evaluation configuration: {"Backbone":"Llama3-8B-Instruct","Method Category":"Prompt-Based"}. This record does not assert independent reproduction.

### Orig.

Factuality Score: 86.2. Result ID: 6995c80f3badfaffe6cbbfb1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-Instruct","Method Category":"Original Model"}. This record does not assert independent reproduction.

### Self-Refine

Factuality Score: 86.2. Result ID: 6995c80f3badfaffe6cbbfb5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f7f1e209131cea45fc4.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-Instruct","Method Category":"Prompt-Based"}. This record does not assert independent reproduction.

### Pair-Few

Factuality Score: 86.1. Result ID: 6995c80f3badfaffe6cbbfb4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-Instruct","Method Category":"Prompt-Based"}. This record does not assert independent reproduction.

### Unc-Few

Factuality Score: 85.6. Result ID: 6995c80f3badfaffe6cbbfb3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

Recorded evaluation configuration: {"Backbone":"Mistral-7B-Instruct","Method Category":"Prompt-Based"}. This record does not assert independent reproduction.

### Orig.

Factuality Score: 85.5. Result ID: 6995c80f3badfaffe6cbbfb8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

Recorded evaluation configuration: {"Backbone":"Llama3-8B-Instruct","Method Category":"Original Model"}. This record does not assert independent reproduction.

### Self-Refine

Factuality Score: 85. Result ID: 6995c80f3badfaffe6cbbfbc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79201e209131cea7c409.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f7f1e209131cea45fc4.md>)

Recorded evaluation configuration: {"Backbone":"Llama3-8B-Instruct","Method Category":"Prompt-Based"}. This record does not assert independent reproduction.

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