# Controllable Language Generation on Word Amazing Pointwise Constraint

Source: [Controllable Language Generation on Word Amazing Pointwise Constraint](<https://www.sota2.com/research/sota/controllable-language-generation-on-word-amazing-pointwise-constraint>)

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

Data updated: 2026-02-26T05:39:49.965Z

Evaluation of language generation models focusing on adherence to pointwise constraints, measured using control metrics (Ctrl) and KL divergence measures against target distributions.

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 | Controllable Language Generation |
| Dataset (including any stated split) | Word Amazing Pointwise Constraint |
| Metric | Control Score |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 5 of 5 |
| 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

- [Control Score](<https://www.sota2.com/research/sota/controllable-language-generation-on-word-amazing-pointwise-constraint.md?metric=Control%20Score>): Higher is better; unit: SCALAR.

- [KL Divergence (p, π)](<https://www.sota2.com/research/sota/controllable-language-generation-on-word-amazing-pointwise-constraint.md?metric=KL%20Divergence%20(p%2C%20%CF%80)>): Lower is better; unit: SCALAR.

- [KL Divergence (p0, π)](<https://www.sota2.com/research/sota/controllable-language-generation-on-word-amazing-pointwise-constraint.md?metric=KL%20Divergence%20(p0%2C%20%CF%80)>): Lower is better; unit: SCALAR.

- [Distribution Distance 1](<https://www.sota2.com/research/sota/controllable-language-generation-on-word-amazing-pointwise-constraint.md?metric=Distribution%20Distance%201>): Higher is better; unit: SCALAR.

- [Distribution Distance 2](<https://www.sota2.com/research/sota/controllable-language-generation-on-word-amazing-pointwise-constraint.md?metric=Distribution%20Distance%202>): Higher is better; unit: SCALAR.

- [Distribution Distance 3](<https://www.sota2.com/research/sota/controllable-language-generation-on-word-amazing-pointwise-constraint.md?metric=Distribution%20Distance%203>): Higher is better; unit: SCALAR.

- [SB Score 4](<https://www.sota2.com/research/sota/controllable-language-generation-on-word-amazing-pointwise-constraint.md?metric=SB%20Score%204>): Higher is better; unit: SCALAR.

- [SB Score 5](<https://www.sota2.com/research/sota/controllable-language-generation-on-word-amazing-pointwise-constraint.md?metric=SB%20Score%205>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Control Score | Result date | Report paper |
| --- | --- | --- | --- |
| Reinforce | 1 | Not recorded | 68fa6dea1e209131cea3ce56 |
| Ziegler | 0.82 | Not recorded | 68fa6dea1e209131cea3ce56 |
| GDC++ | 0.69 | Not recorded | 68fa6dea1e209131cea3ce56 |
| GDC | 0.65 | Not recorded | 68fa6dea1e209131cea3ce56 |
| Original LM | 0 | Not recorded | 68fa6dea1e209131cea3ce56 |

### Reinforce

Control Score: 1. Result ID: 698ffb343a6c314849fd3996.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dea1e209131cea3ce56.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Ziegler

Control Score: 0.82. Result ID: 698ffb343a6c314849fd3997.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dea1e209131cea3ce56.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66e31e209131cea12d84.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### GDC++

Control Score: 0.69. Result ID: 698ffb343a6c314849fd3999.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dea1e209131cea3ce56.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dea1e209131cea3ce56.md>)

[Implementation](<https://github.com/naver/gdc>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### GDC

Control Score: 0.65. Result ID: 698ffb343a6c314849fd3998.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dea1e209131cea3ce56.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69ec1e209131cea24ef3.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Original LM

Control Score: 0. Result ID: 698ffb343a6c314849fd3995.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dea1e209131cea3ce56.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

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