# Instruction Hierarchy Reasoning on MANYIH-BENCH coding

Source: [Instruction Hierarchy Reasoning on MANYIH-BENCH coding](<https://www.sota2.com/research/sota/instruction-hierarchy-reasoning-on-manyih-bench-coding>)

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

Data updated: 2026-04-13T10:02:07.152Z

Evaluation of model performance on instruction hierarchy reasoning capabilities using the MANYIH-BENCH coding dataset.

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 | Instruction Hierarchy Reasoning |
| Dataset (including any stated split) | MANYIH-BENCH coding |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 10 of 10 |
| 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

- [Accuracy](<https://www.sota2.com/research/sota/instruction-hierarchy-reasoning-on-manyih-bench-coding.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [Test Accuracy](<https://www.sota2.com/research/sota/instruction-hierarchy-reasoning-on-manyih-bench-coding.md?metric=Test%20Accuracy>): Higher is better; unit: PERCENT.

- [Style Accuracy](<https://www.sota2.com/research/sota/instruction-hierarchy-reasoning-on-manyih-bench-coding.md?metric=Style%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| GPT 5.4 | 60.9 | Not recorded | 69dc78a27d0172140ec78c04 |
| Gemini 3.1 Pro | 59 | Not recorded | 69dc78a27d0172140ec78c04 |
| Grok 4.20 | 54.1 | Not recorded | 69dc78a27d0172140ec78c04 |
| Opus 4.6 | 51.3 | Not recorded | 69dc78a27d0172140ec78c04 |
| Kimi K2.5 | 42.4 | Not recorded | 69dc78a27d0172140ec78c04 |
| Qwen3.5-397B | 41 | Not recorded | 69dc78a27d0172140ec78c04 |
| Sonnet 4.6 | 39.1 | Not recorded | 69dc78a27d0172140ec78c04 |
| Qwen3.5-122B | 19.7 | Not recorded | 69dc78a27d0172140ec78c04 |
| Qwen3.5-9B | 8.4 | Not recorded | 69dc78a27d0172140ec78c04 |
| Qwen3.5-4B | 3.5 | Not recorded | 69dc78a27d0172140ec78c04 |

### GPT 5.4

Accuracy: 60.9. Result ID: 69dcbf0447f34ce924c42b9a.

[Reporting paper](<https://www.sota2.com/research/paper/69dc78a27d0172140ec78c04.md>)

Recorded evaluation configuration: {"Privilege Prompt Interface":"ordinal"}. This record does not assert independent reproduction.

### Gemini 3.1 Pro

Accuracy: 59. Result ID: 69dcbf0447f34ce924c42b9b.

[Reporting paper](<https://www.sota2.com/research/paper/69dc78a27d0172140ec78c04.md>)

Recorded evaluation configuration: {"Privilege Prompt Interface":"ordinal"}. This record does not assert independent reproduction.

### Grok 4.20

Accuracy: 54.1. Result ID: 69dcbf0447f34ce924c42b9c.

[Reporting paper](<https://www.sota2.com/research/paper/69dc78a27d0172140ec78c04.md>)

Recorded evaluation configuration: {"Privilege Prompt Interface":"ordinal"}. This record does not assert independent reproduction.

### Opus 4.6

Accuracy: 51.3. Result ID: 69dcbf0447f34ce924c42b9d.

[Reporting paper](<https://www.sota2.com/research/paper/69dc78a27d0172140ec78c04.md>)

Recorded evaluation configuration: {"Privilege Prompt Interface":"ordinal"}. This record does not assert independent reproduction.

### Kimi K2.5

Accuracy: 42.4. Result ID: 69dcbf0447f34ce924c42b9e.

[Reporting paper](<https://www.sota2.com/research/paper/69dc78a27d0172140ec78c04.md>)

[Method origin paper](<https://www.sota2.com/research/paper/698181b4a8b01e76cc8e2812.md>)

[Implementation](<https://github.com/MoonshotAI/Kimi-K2.5>)

Recorded evaluation configuration: {"Privilege Prompt Interface":"ordinal"}. This record does not assert independent reproduction.

### Qwen3.5-397B

Accuracy: 41. Result ID: 69dcbf0447f34ce924c42ba0.

[Reporting paper](<https://www.sota2.com/research/paper/69dc78a27d0172140ec78c04.md>)

Recorded evaluation configuration: {"Privilege Prompt Interface":"ordinal"}. This record does not assert independent reproduction.

### Sonnet 4.6

Accuracy: 39.1. Result ID: 69dcbf0447f34ce924c42b9f.

[Reporting paper](<https://www.sota2.com/research/paper/69dc78a27d0172140ec78c04.md>)

Recorded evaluation configuration: {"Privilege Prompt Interface":"ordinal"}. This record does not assert independent reproduction.

### Qwen3.5-122B

Accuracy: 19.7. Result ID: 69dcbf0447f34ce924c42ba1.

[Reporting paper](<https://www.sota2.com/research/paper/69dc78a27d0172140ec78c04.md>)

Recorded evaluation configuration: {"Privilege Prompt Interface":"ordinal"}. This record does not assert independent reproduction.

### Qwen3.5-9B

Accuracy: 8.4. Result ID: 69dcbf0447f34ce924c42ba2.

[Reporting paper](<https://www.sota2.com/research/paper/69dc78a27d0172140ec78c04.md>)

Recorded evaluation configuration: {"Privilege Prompt Interface":"ordinal"}. This record does not assert independent reproduction.

### Qwen3.5-4B

Accuracy: 3.5. Result ID: 69dcbf0447f34ce924c42ba3.

[Reporting paper](<https://www.sota2.com/research/paper/69dc78a27d0172140ec78c04.md>)

Recorded evaluation configuration: {"Privilege Prompt Interface":"ordinal"}. This record does not assert independent reproduction.

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