# Backdoor Detection on SST2 high poison rate (test)

Source: [Backdoor Detection on SST2 high poison rate (test)](<https://www.sota2.com/research/sota/backdoor-detection-on-sst2-high-poison-rate-test>)

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

Data updated: 2026-07-01T07:37:13.579Z

Evaluation of backdoor detection performance on the test split of the SST-2 dataset configured with a high poison rate.

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 | Backdoor Detection |
| Dataset (including any stated split) | SST2 high poison rate (test) |
| Metric | Clean Performance Score |
| Unit | SCALAR |
| 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

- [Clean Performance Score](<https://www.sota2.com/research/sota/backdoor-detection-on-sst2-high-poison-rate-test.md?metric=Clean%20Performance%20Score>): Higher is better; unit: SCALAR.

- [Detection Score (Tell Me Seriously)](<https://www.sota2.com/research/sota/backdoor-detection-on-sst2-high-poison-rate-test.md?metric=Detection%20Score%20(Tell%20Me%20Seriously)>): Higher is better; unit: SCALAR.

- [Detection Metric (mb)](<https://www.sota2.com/research/sota/backdoor-detection-on-sst2-high-poison-rate-test.md?metric=Detection%20Metric%20(mb)>): Higher is better; unit: SCALAR.

- [Detection Score (Tell Me Seriously, Clean)](<https://www.sota2.com/research/sota/backdoor-detection-on-sst2-high-poison-rate-test.md?metric=Detection%20Score%20(Tell%20Me%20Seriously%2C%20Clean)>): Higher is better; unit: SCALAR.

- [Detection Score (Tell Me Seriously, Dirty)](<https://www.sota2.com/research/sota/backdoor-detection-on-sst2-high-poison-rate-test.md?metric=Detection%20Score%20(Tell%20Me%20Seriously%2C%20Dirty)>): Higher is better; unit: SCALAR.

- [Detection Metric (mb, Clean)](<https://www.sota2.com/research/sota/backdoor-detection-on-sst2-high-poison-rate-test.md?metric=Detection%20Metric%20(mb%2C%20Clean)>): Higher is better; unit: SCALAR.

- [Detection Metric (mb, Dirty)](<https://www.sota2.com/research/sota/backdoor-detection-on-sst2-high-poison-rate-test.md?metric=Detection%20Metric%20(mb%2C%20Dirty)>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Clean Performance Score | Result date | Report paper |
| --- | --- | --- | --- |
| MM | 10 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| MM | 10 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| CLIBE | 8 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| CLIBE | 8 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| MM | 8 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| DBS | 8 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| PICCOLO | 6 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| DBS | 4 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| DBS | 4 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| CLIBE | 4 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| CSO(discrete) | 2 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| CSO(continuous) | 1 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| CSO(continuous) | 1 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| CSO(continuous) | 1 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| CSO(discrete) | 0 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| CSO(discrete) | 0 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| PICCOLO | 0 | Not recorded | 6a449ef0e0224bd4f4992df6 |
| PICCOLO | 0 | Not recorded | 6a449ef0e0224bd4f4992df6 |

### MM

Clean Performance Score: 10. Result ID: 6a44c3a873d097d845be8bc7.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70381e209131cea49a5e.md>)

Recorded evaluation configuration: {"Label Type":"clean label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

### MM

Clean Performance Score: 10. Result ID: 6a44c3a873d097d845be8bc8.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70381e209131cea49a5e.md>)

Recorded evaluation configuration: {"Label Type":"dirty label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

### CLIBE

Clean Performance Score: 8. Result ID: 6a44c3a873d097d845be8bcd.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74cf1e209131cea6401c.md>)

Recorded evaluation configuration: {"Label Type":"clean label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

### CLIBE

Clean Performance Score: 8. Result ID: 6a44c3a873d097d845be8bce.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74cf1e209131cea6401c.md>)

Recorded evaluation configuration: {"Label Type":"dirty label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

### MM

Clean Performance Score: 8. Result ID: 6a44c3a873d097d845be8c61.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70381e209131cea49a5e.md>)

Recorded evaluation configuration: {"Model":"Flan-t5-large"}. This record does not assert independent reproduction.

### DBS

Clean Performance Score: 8. Result ID: 6a44c3a873d097d845be8c62.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bc01e209131cea300f7.md>)

Recorded evaluation configuration: {"Model":"Flan-t5-large"}. This record does not assert independent reproduction.

### PICCOLO

Clean Performance Score: 6. Result ID: 6a44c3a873d097d845be8c63.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

Recorded evaluation configuration: {"Model":"Flan-t5-large"}. This record does not assert independent reproduction.

### DBS

Clean Performance Score: 4. Result ID: 6a44c3a873d097d845be8bc9.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bc01e209131cea300f7.md>)

Recorded evaluation configuration: {"Label Type":"clean label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

### DBS

Clean Performance Score: 4. Result ID: 6a44c3a873d097d845be8bca.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bc01e209131cea300f7.md>)

Recorded evaluation configuration: {"Label Type":"dirty label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

### CLIBE

Clean Performance Score: 4. Result ID: 6a44c3a873d097d845be8c64.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74cf1e209131cea6401c.md>)

Recorded evaluation configuration: {"Model":"Flan-t5-large"}. This record does not assert independent reproduction.

### CSO(discrete)

Clean Performance Score: 2. Result ID: 6a44c3a873d097d845be8c60.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

Recorded evaluation configuration: {"Model":"Flan-t5-large"}. This record does not assert independent reproduction.

### CSO(continuous)

Clean Performance Score: 1. Result ID: 6a44c3a873d097d845be8bc3.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

Recorded evaluation configuration: {"Label Type":"clean label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

### CSO(continuous)

Clean Performance Score: 1. Result ID: 6a44c3a873d097d845be8bc4.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

Recorded evaluation configuration: {"Label Type":"dirty label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

### CSO(continuous)

Clean Performance Score: 1. Result ID: 6a44c3a873d097d845be8c5f.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

Recorded evaluation configuration: {"Model":"Flan-t5-large"}. This record does not assert independent reproduction.

### CSO(discrete)

Clean Performance Score: 0. Result ID: 6a44c3a873d097d845be8bc5.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

Recorded evaluation configuration: {"Label Type":"clean label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

### CSO(discrete)

Clean Performance Score: 0. Result ID: 6a44c3a873d097d845be8bc6.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

Recorded evaluation configuration: {"Label Type":"dirty label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

### PICCOLO

Clean Performance Score: 0. Result ID: 6a44c3a873d097d845be8bcb.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

Recorded evaluation configuration: {"Label Type":"clean label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

### PICCOLO

Clean Performance Score: 0. Result ID: 6a44c3a873d097d845be8bcc.

[Reporting paper](<https://www.sota2.com/research/paper/6a449ef0e0224bd4f4992df6.md>)

Recorded evaluation configuration: {"Label Type":"dirty label","Model Architecture":"Qwen3-4b"}. This record does not assert independent reproduction.

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