# Chain-of-Thought Faithfulness Evaluation on Math Book

Source: [Chain-of-Thought Faithfulness Evaluation on Math Book](<https://www.sota2.com/research/sota/chain-of-thought-faithfulness-evaluation-on-math-book>)

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

Data updated: 2026-07-08T07:41:24.681Z

Evaluation of Chain-of-Thought faithfulness on the Math Book dataset, measuring the percentage of unfaithful explanations.

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 | Chain-of-Thought Faithfulness Evaluation |
| Dataset (including any stated split) | Math Book |
| Metric | Unfaithful Explanation Rate |
| Unit | PERCENT |
| Direction | Lower 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

- [Unfaithful Explanation Rate](<https://www.sota2.com/research/sota/chain-of-thought-faithfulness-evaluation-on-math-book.md?metric=Unfaithful%20Explanation%20Rate>): Lower is better; unit: PERCENT.

## Recorded results

| Method | Unfaithful Explanation Rate | Result date | Report paper |
| --- | --- | --- | --- |
| DPO + RMC | 22.8 | Not recorded | 68fa7a0d1e209131cea80d2f |
| DPO + RMC | 23.6 | Not recorded | 68fa7a0d1e209131cea80d2f |
| DPO + RMD | 24.5 | Not recorded | 68fa7a0d1e209131cea80d2f |
| Base LLAMA-3.1-8B-IT | 24.8 | Not recorded | 68fa7a0d1e209131cea80d2f |
| DPO + RM | 25.7 | Not recorded | 68fa7a0d1e209131cea80d2f |
| Base LLAMA-3.1-8B-IT | 27.2 | Not recorded | 68fa7a0d1e209131cea80d2f |
| DPO + RM | 27.2 | Not recorded | 68fa7a0d1e209131cea80d2f |
| DPO + RMD | 28.3 | Not recorded | 68fa7a0d1e209131cea80d2f |
| DPO + RMC | 31.6 | Not recorded | 68fa7a0d1e209131cea80d2f |
| DPO + RMC | 32.5 | Not recorded | 68fa7a0d1e209131cea80d2f |
| DPO + RMD | 33.6 | Not recorded | 68fa7a0d1e209131cea80d2f |
| DPO + RM | 33.9 | Not recorded | 68fa7a0d1e209131cea80d2f |
| DPO + RM | 34 | Not recorded | 68fa7a0d1e209131cea80d2f |
| DPO + RMD | 35.2 | Not recorded | 68fa7a0d1e209131cea80d2f |

### DPO + RMC

Unfaithful Explanation Rate: 22.8. Result ID: 6a4dff1460b88bb339d75b7d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

Recorded evaluation configuration: {"Reward Model":"SK-LLAMA-8B","Decoding":"Greedy"}. This record does not assert independent reproduction.

### DPO + RMC

Unfaithful Explanation Rate: 23.6. Result ID: 6a4dff1460b88bb339d75b83.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

Recorded evaluation configuration: {"Reward Model":"SK-GEMMA-27B","Decoding":"Greedy"}. This record does not assert independent reproduction.

### DPO + RMD

Unfaithful Explanation Rate: 24.5. Result ID: 6a4dff1460b88bb339d75b7b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

Recorded evaluation configuration: {"Reward Model":"SK-LLAMA-8B","Decoding":"Greedy"}. This record does not assert independent reproduction.

### Base LLAMA-3.1-8B-IT

Unfaithful Explanation Rate: 24.8. Result ID: 6a4dff1460b88bb339d75b77.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d11e209131cea6a028.md>)

[Implementation](<https://github.com/meta-llama/llama-agentic-system>)

Recorded evaluation configuration: {"Reward Model":"None","Decoding":"Greedy"}. This record does not assert independent reproduction.

### DPO + RM

Unfaithful Explanation Rate: 25.7. Result ID: 6a4dff1460b88bb339d75b79.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74511e209131cea6105d.md>)

Recorded evaluation configuration: {"Reward Model":"SK-LLAMA-8B","Decoding":"Greedy"}. This record does not assert independent reproduction.

### Base LLAMA-3.1-8B-IT

Unfaithful Explanation Rate: 27.2. Result ID: 6a4dff1460b88bb339d75b78.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d11e209131cea6a028.md>)

[Implementation](<https://github.com/meta-llama/llama-agentic-system>)

Recorded evaluation configuration: {"Reward Model":"None","Decoding":"Maj@16"}. This record does not assert independent reproduction.

### DPO + RM

Unfaithful Explanation Rate: 27.2. Result ID: 6a4dff1460b88bb339d75b7f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74511e209131cea6105d.md>)

Recorded evaluation configuration: {"Reward Model":"SK-GEMMA-27B","Decoding":"Greedy"}. This record does not assert independent reproduction.

### DPO + RMD

Unfaithful Explanation Rate: 28.3. Result ID: 6a4dff1460b88bb339d75b81.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

Recorded evaluation configuration: {"Reward Model":"SK-GEMMA-27B","Decoding":"Greedy"}. This record does not assert independent reproduction.

### DPO + RMC

Unfaithful Explanation Rate: 31.6. Result ID: 6a4dff1460b88bb339d75b7e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

Recorded evaluation configuration: {"Reward Model":"SK-LLAMA-8B","Decoding":"Maj@16"}. This record does not assert independent reproduction.

### DPO + RMC

Unfaithful Explanation Rate: 32.5. Result ID: 6a4dff1460b88bb339d75b84.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

Recorded evaluation configuration: {"Reward Model":"SK-GEMMA-27B","Decoding":"Maj@16"}. This record does not assert independent reproduction.

### DPO + RMD

Unfaithful Explanation Rate: 33.6. Result ID: 6a4dff1460b88bb339d75b7c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

Recorded evaluation configuration: {"Reward Model":"SK-LLAMA-8B","Decoding":"Maj@16"}. This record does not assert independent reproduction.

### DPO + RM

Unfaithful Explanation Rate: 33.9. Result ID: 6a4dff1460b88bb339d75b80.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74511e209131cea6105d.md>)

Recorded evaluation configuration: {"Reward Model":"SK-GEMMA-27B","Decoding":"Maj@16"}. This record does not assert independent reproduction.

### DPO + RM

Unfaithful Explanation Rate: 34. Result ID: 6a4dff1460b88bb339d75b7a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74511e209131cea6105d.md>)

Recorded evaluation configuration: {"Reward Model":"SK-LLAMA-8B","Decoding":"Maj@16"}. This record does not assert independent reproduction.

### DPO + RMD

Unfaithful Explanation Rate: 35.2. Result ID: 6a4dff1460b88bb339d75b82.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a0d1e209131cea80d2f.md>)

Recorded evaluation configuration: {"Reward Model":"SK-GEMMA-27B","Decoding":"Maj@16"}. This record does not assert independent reproduction.

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