# Human Behavior Modeling on CVABench (test)

Source: [Human Behavior Modeling on CVABench (test)](<https://www.sota2.com/research/sota/human-behavior-modeling-on-cvabench-test>)

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

Data updated: 2026-04-08T09:31:56.631Z

Evaluation of model performance on the test split of the CVABench dataset, assessing human behavior modeling capabilities across linguistic consistency, sentiment, conversation, and travel domains.

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 | Human Behavior Modeling |
| Dataset (including any stated split) | CVABench (test) |
| Metric | Media Linguistic TTR |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 7 of 7 |
| 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

- [Media Linguistic TTR](<https://www.sota2.com/research/sota/human-behavior-modeling-on-cvabench-test.md?metric=Media%20Linguistic%20TTR>): Higher is better; unit: SCALAR.

- [Media Rate Accuracy](<https://www.sota2.com/research/sota/human-behavior-modeling-on-cvabench-test.md?metric=Media%20Rate%20Accuracy>): Higher is better; unit: PERCENT.

- [Media Sentiment Accuracy](<https://www.sota2.com/research/sota/human-behavior-modeling-on-cvabench-test.md?metric=Media%20Sentiment%20Accuracy>): Higher is better; unit: PERCENT.

- [Conversation Linguistic TTR](<https://www.sota2.com/research/sota/human-behavior-modeling-on-cvabench-test.md?metric=Conversation%20Linguistic%20TTR>): Higher is better; unit: SCALAR.

- [Conversation Sentiment Accuracy](<https://www.sota2.com/research/sota/human-behavior-modeling-on-cvabench-test.md?metric=Conversation%20Sentiment%20Accuracy>): Higher is better; unit: PERCENT.

- [Travel Polarity Accuracy](<https://www.sota2.com/research/sota/human-behavior-modeling-on-cvabench-test.md?metric=Travel%20Polarity%20Accuracy>): Higher is better; unit: PERCENT.

- [Travel Stay MSE](<https://www.sota2.com/research/sota/human-behavior-modeling-on-cvabench-test.md?metric=Travel%20Stay%20MSE>): Lower is better; unit: SCALAR.

- [Overall Value Change (%)](<https://www.sota2.com/research/sota/human-behavior-modeling-on-cvabench-test.md?metric=Overall%20Value%20Change%20(%25)>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Media Linguistic TTR | Result date | Report paper |
| --- | --- | --- | --- |
| Qwen2.5-7B-SFT+DPO | 0.07 | Not recorded | 69d5e13e7d0172140ec75c9f |
| Role-Play Agent | 0.06 | Not recorded | 69d5e13e7d0172140ec75c9f |
| Reasoning Agent - 0 | 0.06 | Not recorded | 69d5e13e7d0172140ec75c9f |
| Reasoning Agent - 2 | 0.06 | Not recorded | 69d5e13e7d0172140ec75c9f |
| Reasoning Agent - 4 | 0.06 | Not recorded | 69d5e13e7d0172140ec75c9f |
| Qwen2.5-7B SFT | 0.06 | Not recorded | 69d5e13e7d0172140ec75c9f |
| CVA (Ours) | 0.04 | Not recorded | 69d5e13e7d0172140ec75c9f |

### Qwen2.5-7B-SFT+DPO

Media Linguistic TTR: 0.07. Result ID: 69d6207a47f34ce924bf079c.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e13e7d0172140ec75c9f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa713c1e209131cea4ff94.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B","Training Mode":"SFT+DPO"}. This record does not assert independent reproduction.

### Role-Play Agent

Media Linguistic TTR: 0.06. Result ID: 69d6207a47f34ce924bf0797.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e13e7d0172140ec75c9f.md>)

Recorded evaluation configuration: {"Backbone":"GPT-4o-mini","Training Mode":"Training-free"}. This record does not assert independent reproduction.

### Reasoning Agent - 0

Media Linguistic TTR: 0.06. Result ID: 69d6207a47f34ce924bf0798.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e13e7d0172140ec75c9f.md>)

Recorded evaluation configuration: {"Backbone":"GPT-4o-mini","Training Mode":"Training-free"}. This record does not assert independent reproduction.

### Reasoning Agent - 2

Media Linguistic TTR: 0.06. Result ID: 69d6207a47f34ce924bf0799.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e13e7d0172140ec75c9f.md>)

Recorded evaluation configuration: {"Backbone":"GPT-4o-mini","Training Mode":"Training-free"}. This record does not assert independent reproduction.

### Reasoning Agent - 4

Media Linguistic TTR: 0.06. Result ID: 69d6207a47f34ce924bf079a.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e13e7d0172140ec75c9f.md>)

Recorded evaluation configuration: {"Backbone":"GPT-4o-mini","Training Mode":"Training-free"}. This record does not assert independent reproduction.

### Qwen2.5-7B SFT

Media Linguistic TTR: 0.06. Result ID: 69d6207a47f34ce924bf079b.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e13e7d0172140ec75c9f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bf71e209131cea314c0.md>)

[Implementation](<https://github.com/openai/following-instructions-human-feedback>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B","Training Mode":"SFT"}. This record does not assert independent reproduction.

### CVA (Ours)

Media Linguistic TTR: 0.04. Result ID: 69d6207a47f34ce924bf079d.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e13e7d0172140ec75c9f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e13e7d0172140ec75c9f.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B"}. This record does not assert independent reproduction.

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