# Dialogue Generation on PERSUASIVETOM

Source: [Dialogue Generation on PERSUASIVETOM](<https://www.sota2.com/research/sota/dialogue-generation-on-persuasivetom>)

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

Data updated: 2026-04-14T07:50:44.327Z

Evaluation of dialogue generation performance on the PERSUASIVETOM dataset, assessed using Theory of Mind (ToM), Coherence (Coh.), and Perspective Taking (PSE) scores derived from Llama-3.3-70B, GPT-5.1, and Human evaluations.

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 | Dialogue Generation |
| Dataset (including any stated split) | PERSUASIVETOM |
| Metric | ToM Score (Judge: Llama-3.3-70B) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 12 of 12 |
| 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

- [ToM Score (Judge: Llama-3.3-70B)](<https://www.sota2.com/research/sota/dialogue-generation-on-persuasivetom.md?metric=ToM%20Score%20(Judge%3A%20Llama-3.3-70B)>): Higher is better; unit: PERCENT.

- [Coherence Score (Judge: Llama-3.3-70B)](<https://www.sota2.com/research/sota/dialogue-generation-on-persuasivetom.md?metric=Coherence%20Score%20(Judge%3A%20Llama-3.3-70B)>): Higher is better; unit: PERCENT.

- [PSE Score (Judge: Llama-3.3-70B)](<https://www.sota2.com/research/sota/dialogue-generation-on-persuasivetom.md?metric=PSE%20Score%20(Judge%3A%20Llama-3.3-70B)>): Higher is better; unit: PERCENT.

- [ToM Score (Judge: GPT-5.1)](<https://www.sota2.com/research/sota/dialogue-generation-on-persuasivetom.md?metric=ToM%20Score%20(Judge%3A%20GPT-5.1)>): Higher is better; unit: PERCENT.

- [Coherence Score (Judge: GPT-5.1)](<https://www.sota2.com/research/sota/dialogue-generation-on-persuasivetom.md?metric=Coherence%20Score%20(Judge%3A%20GPT-5.1)>): Higher is better; unit: PERCENT.

- [PSE Score (Judge: GPT-5.1)](<https://www.sota2.com/research/sota/dialogue-generation-on-persuasivetom.md?metric=PSE%20Score%20(Judge%3A%20GPT-5.1)>): Higher is better; unit: PERCENT.

- [ToM Score (Judge: Human)](<https://www.sota2.com/research/sota/dialogue-generation-on-persuasivetom.md?metric=ToM%20Score%20(Judge%3A%20Human)>): Higher is better; unit: PERCENT.

- [Coherence Score (Judge: Human)](<https://www.sota2.com/research/sota/dialogue-generation-on-persuasivetom.md?metric=Coherence%20Score%20(Judge%3A%20Human)>): Higher is better; unit: PERCENT.

- [PSE Score (Judge: Human)](<https://www.sota2.com/research/sota/dialogue-generation-on-persuasivetom.md?metric=PSE%20Score%20(Judge%3A%20Human)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ToM Score (Judge: Llama-3.3-70B) | Result date | Report paper |
| --- | --- | --- | --- |
| COSTOM | 80.2 | Not recorded | 69ddc9ed7d0172140ec79860 |
| COSTOM | 79.7 | Not recorded | 69ddc9ed7d0172140ec79860 |
| LatentQA | 67.1 | Not recorded | 69ddc9ed7d0172140ec79860 |
| Full-Layer LoRA | 65.4 | Not recorded | 69ddc9ed7d0172140ec79860 |
| MindDial (Fine-tuned) | 64.3 | Not recorded | 69ddc9ed7d0172140ec79860 |
| MindDial (Fine-tuned) | 64.3 | Not recorded | 69ddc9ed7d0172140ec79860 |
| Full-Layer LoRA | 62.8 | Not recorded | 69ddc9ed7d0172140ec79860 |
| MindDial (prompt) | 45.1 | Not recorded | 69ddc9ed7d0172140ec79860 |
| MindDial (prompt) | 44.2 | Not recorded | 69ddc9ed7d0172140ec79860 |
| LatentQA | 25.5 | Not recorded | 69ddc9ed7d0172140ec79860 |
| Zero-shot Baseline | 10.9 | Not recorded | 69ddc9ed7d0172140ec79860 |
| Zero-shot Baseline | 10.4 | Not recorded | 69ddc9ed7d0172140ec79860 |

### COSTOM

ToM Score (Judge: Llama-3.3-70B): 80.2. Result ID: 69ddf18d47f34ce924c51d25.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Qwen2.5-7B-Instruct","Category":"Tuning & Intervention"}. This record does not assert independent reproduction.

### COSTOM

ToM Score (Judge: Llama-3.3-70B): 79.7. Result ID: 69ddf18d47f34ce924c51d1f.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Llama-3-8B-Instruct","Category":"Tuning & Intervention"}. This record does not assert independent reproduction.

### LatentQA

ToM Score (Judge: Llama-3.3-70B): 67.1. Result ID: 69ddf18d47f34ce924c51d24.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Qwen2.5-7B-Instruct","Category":"Tuning & Intervention"}. This record does not assert independent reproduction.

### Full-Layer LoRA

ToM Score (Judge: Llama-3.3-70B): 65.4. Result ID: 69ddf18d47f34ce924c51d23.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Qwen2.5-7B-Instruct","Category":"Tuning & Intervention"}. This record does not assert independent reproduction.

### MindDial (Fine-tuned)

ToM Score (Judge: Llama-3.3-70B): 64.3. Result ID: 69ddf18d47f34ce924c51d1c.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Llama-3-8B-Instruct","Category":"Tuning & Intervention"}. This record does not assert independent reproduction.

### MindDial (Fine-tuned)

ToM Score (Judge: Llama-3.3-70B): 64.3. Result ID: 69ddf18d47f34ce924c51d22.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Qwen2.5-7B-Instruct","Category":"Tuning & Intervention"}. This record does not assert independent reproduction.

### Full-Layer LoRA

ToM Score (Judge: Llama-3.3-70B): 62.8. Result ID: 69ddf18d47f34ce924c51d1d.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Llama-3-8B-Instruct","Category":"Tuning & Intervention"}. This record does not assert independent reproduction.

### MindDial (prompt)

ToM Score (Judge: Llama-3.3-70B): 45.1. Result ID: 69ddf18d47f34ce924c51d1b.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Llama-3-8B-Instruct","Category":"Prompting Baselines"}. This record does not assert independent reproduction.

### MindDial (prompt)

ToM Score (Judge: Llama-3.3-70B): 44.2. Result ID: 69ddf18d47f34ce924c51d21.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Qwen2.5-7B-Instruct","Category":"Prompting Baselines"}. This record does not assert independent reproduction.

### LatentQA

ToM Score (Judge: Llama-3.3-70B): 25.5. Result ID: 69ddf18d47f34ce924c51d1e.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Llama-3-8B-Instruct","Category":"Tuning & Intervention"}. This record does not assert independent reproduction.

### Zero-shot Baseline

ToM Score (Judge: Llama-3.3-70B): 10.9. Result ID: 69ddf18d47f34ce924c51d20.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Qwen2.5-7B-Instruct","Category":"Prompting Baselines"}. This record does not assert independent reproduction.

### Zero-shot Baseline

ToM Score (Judge: Llama-3.3-70B): 10.4. Result ID: 69ddf18d47f34ce924c51d1a.

[Reporting paper](<https://www.sota2.com/research/paper/69ddc9ed7d0172140ec79860.md>)

Recorded evaluation configuration: {"Base Model":"Llama-3-8B-Instruct","Category":"Prompting Baselines"}. This record does not assert independent reproduction.

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