# Zero-shot cross-task generalization on Super-NaturalInstructions (test)

Source: [Zero-shot cross-task generalization on Super-NaturalInstructions (test)](<https://www.sota2.com/research/sota/zero-shot-cross-task-generalization-on-super-naturalinstructions-test>)

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

Data updated: 2026-02-26T06:00:42.003Z

Evaluation of zero-shot cross-task generalization performance using the test split of the Super-NaturalInstructions dataset, typically measured using ROUGE metrics.

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 | Zero-shot cross-task generalization |
| Dataset (including any stated split) | Super-NaturalInstructions (test) |
| Metric | ROUGE-L |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 8 of 8 |
| 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

- [ROUGE-L](<https://www.sota2.com/research/sota/zero-shot-cross-task-generalization-on-super-naturalinstructions-test.md?metric=ROUGE-L>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ROUGE-L | Result date | Report paper |
| --- | --- | --- | --- |
| Def. + Instruction Embedder | 34.97 | Not recorded | 68fa6fcd1e209131cea477f7 |
| Def. + Instruction Embedder | 34.79 | Not recorded | 68fa6fcd1e209131cea477f7 |
| Def. | 33.82 | Not recorded | 68fa6fcd1e209131cea477f7 |
| Def. + Pos. + Neg. + Expl. | 29.07 | Not recorded | 68fa6fcd1e209131cea477f7 |
| Def. + Pos. + Neg. | 27.91 | Not recorded | 68fa6fcd1e209131cea477f7 |
| Def. + Pos. | 27.74 | Not recorded | 68fa6fcd1e209131cea477f7 |
| Instruction Embedder | 11.99 | Not recorded | 68fa6fcd1e209131cea477f7 |
| Instruction Embedder | 11.79 | Not recorded | 68fa6fcd1e209131cea477f7 |

### Def. + Instruction Embedder

ROUGE-L: 34.97. Result ID: 6990fede3a6c31484904e1e3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Implementation](<https://github.com/misonuma/instopt>)

Recorded evaluation configuration: {"Optimization":"Instance Conversion (IC)","Training Instruction":"Def. + learnable instruction"}. This record does not assert independent reproduction.

### Def. + Instruction Embedder

ROUGE-L: 34.79. Result ID: 6990fede3a6c31484904e1e2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Implementation](<https://github.com/misonuma/instopt>)

Recorded evaluation configuration: {"Optimization":"Direct Parameterization (DP)","Training Instruction":"Def. + learnable instruction"}. This record does not assert independent reproduction.

### Def.

ROUGE-L: 33.82. Result ID: 6990fede3a6c31484904e1dc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dc11e209131cea3bdf1.md>)

Recorded evaluation configuration: {"Training Instruction":"Task definition"}. This record does not assert independent reproduction.

### Def. + Pos. + Neg. + Expl.

ROUGE-L: 29.07. Result ID: 6990fede3a6c31484904e1df.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dc11e209131cea3bdf1.md>)

Recorded evaluation configuration: {"Training Instruction":"Task definition + 1 positive/negative exemplar + explanation"}. This record does not assert independent reproduction.

### Def. + Pos. + Neg.

ROUGE-L: 27.91. Result ID: 6990fede3a6c31484904e1de.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dc11e209131cea3bdf1.md>)

Recorded evaluation configuration: {"Training Instruction":"Task definition + 1 positive exemplar + 1 negative exemplar"}. This record does not assert independent reproduction.

### Def. + Pos.

ROUGE-L: 27.74. Result ID: 6990fede3a6c31484904e1dd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dc11e209131cea3bdf1.md>)

Recorded evaluation configuration: {"Training Instruction":"Task definition + 1 positive exemplar"}. This record does not assert independent reproduction.

### Instruction Embedder

ROUGE-L: 11.99. Result ID: 6990fede3a6c31484904e1e1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Implementation](<https://github.com/misonuma/instopt>)

Recorded evaluation configuration: {"Optimization":"Instance Conversion (IC)"}. This record does not assert independent reproduction.

### Instruction Embedder

ROUGE-L: 11.79. Result ID: 6990fede3a6c31484904e1e0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea477f7.md>)

[Implementation](<https://github.com/misonuma/instopt>)

Recorded evaluation configuration: {"Optimization":"Direct Parameterization (DP)"}. This record does not assert independent reproduction.

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