# Computer Use Control-Flow Hijacking on CFH-Hard Computer Use

Source: [Computer Use Control-Flow Hijacking on CFH-Hard Computer Use](<https://www.sota2.com/research/sota/computer-use-control-flow-hijacking-on-cfh-hard-computer-use>)

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

Data updated: 2026-03-06T08:48:48.703Z

Evaluation of model performance on the CFH-Hard Computer Use dataset, focusing on control-flow hijacking vulnerabilities, measured by Gen., Py., WP, FNF, and ANF rates.

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 | Computer Use Control-Flow Hijacking |
| Dataset (including any stated split) | CFH-Hard Computer Use |
| Metric | Gen. Rate |
| 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

- [Gen. Rate](<https://www.sota2.com/research/sota/computer-use-control-flow-hijacking-on-cfh-hard-computer-use.md?metric=Gen.%20Rate>): Higher is better; unit: PERCENT.

- [Py. Rate](<https://www.sota2.com/research/sota/computer-use-control-flow-hijacking-on-cfh-hard-computer-use.md?metric=Py.%20Rate>): Higher is better; unit: PERCENT.

- [WP Rate](<https://www.sota2.com/research/sota/computer-use-control-flow-hijacking-on-cfh-hard-computer-use.md?metric=WP%20Rate>): Higher is better; unit: PERCENT.

- [FNF Rate](<https://www.sota2.com/research/sota/computer-use-control-flow-hijacking-on-cfh-hard-computer-use.md?metric=FNF%20Rate>): Higher is better; unit: PERCENT.

- [ANF Rate](<https://www.sota2.com/research/sota/computer-use-control-flow-hijacking-on-cfh-hard-computer-use.md?metric=ANF%20Rate>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Gen. Rate | Result date | Report paper |
| --- | --- | --- | --- |
| Undefended | 67 | Not recorded | 68fa7c591e209131cea8db33 |
| ACF | 67 | Not recorded | 68fa7c591e209131cea8db33 |
| LP | 67 | Not recorded | 68fa7c591e209131cea8db33 |
| LlamaFirewall | 67 | Not recorded | 68fa7c591e209131cea8db33 |
| LlamaFirewall | 67 | Not recorded | 68fa7c591e209131cea8db33 |
| LlamaFirewall | 56 | Not recorded | 68fa7c591e209131cea8db33 |
| LlamaFirewall | 6 | Not recorded | 68fa7c591e209131cea8db33 |
| CONTROLVALVE | 0 | Not recorded | 68fa7c591e209131cea8db33 |

### Undefended

Gen. Rate: 67. Result ID: 69aa94cf6dbb7075b9546d3a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c591e209131cea8db33.md>)

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

### ACF

Gen. Rate: 67. Result ID: 69aa94cf6dbb7075b9546d3b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c591e209131cea8db33.md>)

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

### LP

Gen. Rate: 67. Result ID: 69aa94cf6dbb7075b9546d3c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c591e209131cea8db33.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7af71e209131cea8604a.md>)

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

### LlamaFirewall

Gen. Rate: 67. Result ID: 69aa94cf6dbb7075b9546d3d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c591e209131cea8db33.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa786c1e209131cea785e3.md>)

Recorded evaluation configuration: {"Backbone":"llama"}. This record does not assert independent reproduction.

### LlamaFirewall

Gen. Rate: 67. Result ID: 69aa94cf6dbb7075b9546d3f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c591e209131cea8db33.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa786c1e209131cea785e3.md>)

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

### LlamaFirewall

Gen. Rate: 56. Result ID: 69aa94cf6dbb7075b9546d3e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c591e209131cea8db33.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa786c1e209131cea785e3.md>)

Recorded evaluation configuration: {"Backbone":"4o"}. This record does not assert independent reproduction.

### LlamaFirewall

Gen. Rate: 6. Result ID: 69aa94cf6dbb7075b9546d40.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c591e209131cea8db33.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa786c1e209131cea785e3.md>)

Recorded evaluation configuration: {"Backbone":"o4-mini"}. This record does not assert independent reproduction.

### CONTROLVALVE

Gen. Rate: 0. Result ID: 69aa94cf6dbb7075b9546d41.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c591e209131cea8db33.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c591e209131cea8db33.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.