# Recall-Intensive Evaluation on Recall-Intensive Tasks

Source: [Recall-Intensive Evaluation on Recall-Intensive Tasks](<https://www.sota2.com/research/sota/recall-intensive-evaluation-on-recall-intensive-tasks>)

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

Data updated: 2026-04-16T07:03:08.109Z

Evaluation measuring accuracy across several recall-intensive QA metrics including FDA, SWDE, TQA, NQ, SQD, and Drop.

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 | Recall-Intensive Evaluation |
| Dataset (including any stated split) | Recall-Intensive Tasks |
| Metric | FDA Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 18 of 18 |
| 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

- [FDA Accuracy](<https://www.sota2.com/research/sota/recall-intensive-evaluation-on-recall-intensive-tasks.md?metric=FDA%20Accuracy>): Higher is better; unit: PERCENT.

- [SWDE Accuracy](<https://www.sota2.com/research/sota/recall-intensive-evaluation-on-recall-intensive-tasks.md?metric=SWDE%20Accuracy>): Higher is better; unit: PERCENT.

- [TQA Accuracy](<https://www.sota2.com/research/sota/recall-intensive-evaluation-on-recall-intensive-tasks.md?metric=TQA%20Accuracy>): Higher is better; unit: PERCENT.

- [NQ Accuracy](<https://www.sota2.com/research/sota/recall-intensive-evaluation-on-recall-intensive-tasks.md?metric=NQ%20Accuracy>): Higher is better; unit: PERCENT.

- [SQD Accuracy](<https://www.sota2.com/research/sota/recall-intensive-evaluation-on-recall-intensive-tasks.md?metric=SQD%20Accuracy>): Higher is better; unit: PERCENT.

- [Drop Accuracy](<https://www.sota2.com/research/sota/recall-intensive-evaluation-on-recall-intensive-tasks.md?metric=Drop%20Accuracy>): Higher is better; unit: PERCENT.

- [Average Accuracy](<https://www.sota2.com/research/sota/recall-intensive-evaluation-on-recall-intensive-tasks.md?metric=Average%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | FDA Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| GLA-H | 69.48 | Not recorded | 68fa7c101e209131cea8c392 |
| GSA-H | 68.3 | Not recorded | 68fa7c101e209131cea8c392 |
| NHA | 68.3 | Not recorded | 68fa7c101e209131cea8c392 |
| GDN-H | 66.76 | Not recorded | 68fa7c101e209131cea8c392 |
| Mamba2-H | 65.3 | Not recorded | 68fa7c101e209131cea8c392 |
| GSA-H | 62.13 | Not recorded | 68fa7c101e209131cea8c392 |
| GLA-H | 58.86 | Not recorded | 68fa7c101e209131cea8c392 |
| Mamba2-H | 55.68 | Not recorded | 68fa7c101e209131cea8c392 |
| NHA | 53.86 | Not recorded | 68fa7c101e209131cea8c392 |
| GDN-H | 52.32 | Not recorded | 68fa7c101e209131cea8c392 |
| Trans++ | 46.14 | Not recorded | 68fa7c101e209131cea8c392 |
| Trans++ | 44.32 | Not recorded | 68fa7c101e209131cea8c392 |
| GDN | 30.25 | Not recorded | 68fa7c101e209131cea8c392 |
| GLA | 27.61 | Not recorded | 68fa7c101e209131cea8c392 |
| GSA | 23.25 | Not recorded | 68fa7c101e209131cea8c392 |
| GDN | 20.53 | Not recorded | 68fa7c101e209131cea8c392 |
| GLA | 11.26 | Not recorded | 68fa7c101e209131cea8c392 |
| GSA | 6.36 | Not recorded | 68fa7c101e209131cea8c392 |

### GLA-H

FDA Accuracy: 69.48. Result ID: 69e0898a47f34ce924c94d8a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74a51e209131cea62e34.md>)

Recorded evaluation configuration: {"Parameters":"1.3B","Training Tokens":"100B","Architecture Type":"Hybrid"}. This record does not assert independent reproduction.

### GSA-H

FDA Accuracy: 68.3. Result ID: 69e0898a47f34ce924c94d8b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa757b1e209131cea68452.md>)

[Implementation](<https://github.com/sustcsonglin/flash-linear-attention>)

Recorded evaluation configuration: {"Parameters":"1.3B","Training Tokens":"100B","Architecture Type":"Hybrid"}. This record does not assert independent reproduction.

### NHA

FDA Accuracy: 68.3. Result ID: 69e0898a47f34ce924c94d8d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

Recorded evaluation configuration: {"Parameters":"1.3B","Training Tokens":"100B","Architecture Type":"Hybrid"}. This record does not assert independent reproduction.

### GDN-H

FDA Accuracy: 66.76. Result ID: 69e0898a47f34ce924c94d8c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa774b1e209131cea72592.md>)

Recorded evaluation configuration: {"Parameters":"1.3B","Training Tokens":"100B","Architecture Type":"Hybrid"}. This record does not assert independent reproduction.

### Mamba2-H

FDA Accuracy: 65.3. Result ID: 69e0898a47f34ce924c94d89.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73641e209131cea5c239.md>)

Recorded evaluation configuration: {"Parameters":"1.3B","Training Tokens":"100B","Architecture Type":"Hybrid"}. This record does not assert independent reproduction.

### GSA-H

FDA Accuracy: 62.13. Result ID: 69e0898a47f34ce924c94d82.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa757b1e209131cea68452.md>)

[Implementation](<https://github.com/sustcsonglin/flash-linear-attention>)

Recorded evaluation configuration: {"Parameters":"340M","Training Tokens":"15B","Architecture Type":"Hybrid"}. This record does not assert independent reproduction.

### GLA-H

FDA Accuracy: 58.86. Result ID: 69e0898a47f34ce924c94d81.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74a51e209131cea62e34.md>)

Recorded evaluation configuration: {"Parameters":"340M","Training Tokens":"15B","Architecture Type":"Hybrid"}. This record does not assert independent reproduction.

### Mamba2-H

FDA Accuracy: 55.68. Result ID: 69e0898a47f34ce924c94d80.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73641e209131cea5c239.md>)

Recorded evaluation configuration: {"Parameters":"340M","Training Tokens":"15B","Architecture Type":"Hybrid"}. This record does not assert independent reproduction.

### NHA

FDA Accuracy: 53.86. Result ID: 69e0898a47f34ce924c94d84.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

Recorded evaluation configuration: {"Parameters":"340M","Training Tokens":"15B","Architecture Type":"Hybrid"}. This record does not assert independent reproduction.

### GDN-H

FDA Accuracy: 52.32. Result ID: 69e0898a47f34ce924c94d83.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa774b1e209131cea72592.md>)

Recorded evaluation configuration: {"Parameters":"340M","Training Tokens":"15B","Architecture Type":"Hybrid"}. This record does not assert independent reproduction.

### Trans++

FDA Accuracy: 46.14. Result ID: 69e0898a47f34ce924c94d7c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa703d1e209131cea49db1.md>)

[Implementation](<https://github.com/tensorflow/tensor2tensor>)

Recorded evaluation configuration: {"Parameters":"340M","Training Tokens":"15B","Architecture Type":"Base"}. This record does not assert independent reproduction.

### Trans++

FDA Accuracy: 44.32. Result ID: 69e0898a47f34ce924c94d85.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa703d1e209131cea49db1.md>)

[Implementation](<https://github.com/tensorflow/tensor2tensor>)

Recorded evaluation configuration: {"Parameters":"1.3B","Training Tokens":"100B","Architecture Type":"Base"}. This record does not assert independent reproduction.

### GDN

FDA Accuracy: 30.25. Result ID: 69e0898a47f34ce924c94d88.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa774b1e209131cea72592.md>)

Recorded evaluation configuration: {"Parameters":"1.3B","Training Tokens":"100B","Architecture Type":"Base"}. This record does not assert independent reproduction.

### GLA

FDA Accuracy: 27.61. Result ID: 69e0898a47f34ce924c94d86.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74a51e209131cea62e34.md>)

Recorded evaluation configuration: {"Parameters":"1.3B","Training Tokens":"100B","Architecture Type":"Base"}. This record does not assert independent reproduction.

### GSA

FDA Accuracy: 23.25. Result ID: 69e0898a47f34ce924c94d87.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa757b1e209131cea68452.md>)

[Implementation](<https://github.com/sustcsonglin/flash-linear-attention>)

Recorded evaluation configuration: {"Parameters":"1.3B","Training Tokens":"100B","Architecture Type":"Base"}. This record does not assert independent reproduction.

### GDN

FDA Accuracy: 20.53. Result ID: 69e0898a47f34ce924c94d7f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa774b1e209131cea72592.md>)

Recorded evaluation configuration: {"Parameters":"340M","Training Tokens":"15B","Architecture Type":"Base"}. This record does not assert independent reproduction.

### GLA

FDA Accuracy: 11.26. Result ID: 69e0898a47f34ce924c94d7d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74a51e209131cea62e34.md>)

Recorded evaluation configuration: {"Parameters":"340M","Training Tokens":"15B","Architecture Type":"Base"}. This record does not assert independent reproduction.

### GSA

FDA Accuracy: 6.36. Result ID: 69e0898a47f34ce924c94d7e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c101e209131cea8c392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa757b1e209131cea68452.md>)

[Implementation](<https://github.com/sustcsonglin/flash-linear-attention>)

Recorded evaluation configuration: {"Parameters":"340M","Training Tokens":"15B","Architecture Type":"Base"}. This record does not assert independent reproduction.

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