# Chunking Strategy Evaluation for RAG on QASC Evaluation Set (5-fold cross-validation)

Source: [Chunking Strategy Evaluation for RAG on QASC Evaluation Set (5-fold cross-validation)](<https://www.sota2.com/research/sota/chunking-strategy-evaluation-for-rag-on-qasc-evaluation-set-5-fold-cross-validation>)

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

Data updated: 2026-05-25T07:17:08.677Z

Evaluation of different chunking strategies for Retrieval-Augmented Generation (RAG) systems using the QASC dataset, measured by metrics including Precision, Recall, F1-Score, Answer Correctness, and Faithfulness.

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 | Chunking Strategy Evaluation for RAG |
| Dataset (including any stated split) | QASC Evaluation Set (5-fold cross-validation) |
| Metric | Precision |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 9 of 9 |
| 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

- [Precision](<https://www.sota2.com/research/sota/chunking-strategy-evaluation-for-rag-on-qasc-evaluation-set-5-fold-cross-validation.md?metric=Precision>): Higher is better; unit: PERCENT.

- [Recall](<https://www.sota2.com/research/sota/chunking-strategy-evaluation-for-rag-on-qasc-evaluation-set-5-fold-cross-validation.md?metric=Recall>): Higher is better; unit: PERCENT.

- [F1-Score](<https://www.sota2.com/research/sota/chunking-strategy-evaluation-for-rag-on-qasc-evaluation-set-5-fold-cross-validation.md?metric=F1-Score>): Higher is better; unit: PERCENT.

- [Answer Correctness](<https://www.sota2.com/research/sota/chunking-strategy-evaluation-for-rag-on-qasc-evaluation-set-5-fold-cross-validation.md?metric=Answer%20Correctness>): Higher is better; unit: PERCENT.

- [Faithfulness](<https://www.sota2.com/research/sota/chunking-strategy-evaluation-for-rag-on-qasc-evaluation-set-5-fold-cross-validation.md?metric=Faithfulness>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Precision | Result date | Report paper |
| --- | --- | --- | --- |
| QASC | 85 | Not recorded | 6a13d77e6f71c5ae7dc9e7e2 |
| Semantic Chunking | 78 | Not recorded | 6a13d77e6f71c5ae7dc9e7e2 |
| Agentic Chunking | 76 | Not recorded | 6a13d77e6f71c5ae7dc9e7e2 |
| Fixed | 74 | Not recorded | 6a13d77e6f71c5ae7dc9e7e2 |
| Recursive Splitting | 74 | Not recorded | 6a13d77e6f71c5ae7dc9e7e2 |
| Fixed | 73 | Not recorded | 6a13d77e6f71c5ae7dc9e7e2 |
| Fixed | 72 | Not recorded | 6a13d77e6f71c5ae7dc9e7e2 |
| Fixed | 71 | Not recorded | 6a13d77e6f71c5ae7dc9e7e2 |
| Fixed | 70 | Not recorded | 6a13d77e6f71c5ae7dc9e7e2 |

### QASC

Precision: 85. Result ID: 6a13f755fbc373592c0ab64e.

[Reporting paper](<https://www.sota2.com/research/paper/6a13d77e6f71c5ae7dc9e7e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a13d77e6f71c5ae7dc9e7e2.md>)

Recorded evaluation configuration: {"Latency (ms)":"380 ± 50"}. This record does not assert independent reproduction.

### Semantic Chunking

Precision: 78. Result ID: 6a13f755fbc373592c0ab64c.

[Reporting paper](<https://www.sota2.com/research/paper/6a13d77e6f71c5ae7dc9e7e2.md>)

Recorded evaluation configuration: {"Latency (ms)":"340 ± 45"}. This record does not assert independent reproduction.

### Agentic Chunking

Precision: 76. Result ID: 6a13f755fbc373592c0ab64d.

[Reporting paper](<https://www.sota2.com/research/paper/6a13d77e6f71c5ae7dc9e7e2.md>)

Recorded evaluation configuration: {"Latency (ms)":"2,850 ± 620"}. This record does not assert independent reproduction.

### Fixed

Precision: 74. Result ID: 6a13f755fbc373592c0ab648.

[Reporting paper](<https://www.sota2.com/research/paper/6a13d77e6f71c5ae7dc9e7e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa709e1e209131cea4c470.md>)

[Implementation](<https://github.com/google-research/text-to-text-transfer-transformer>)

Recorded evaluation configuration: {"Chunk Size":"500","Latency (ms)":"110 ± 10"}. This record does not assert independent reproduction.

### Recursive Splitting

Precision: 74. Result ID: 6a13f755fbc373592c0ab64b.

[Reporting paper](<https://www.sota2.com/research/paper/6a13d77e6f71c5ae7dc9e7e2.md>)

Recorded evaluation configuration: {"Latency (ms)":"118 ± 13"}. This record does not assert independent reproduction.

### Fixed

Precision: 73. Result ID: 6a13f755fbc373592c0ab649.

[Reporting paper](<https://www.sota2.com/research/paper/6a13d77e6f71c5ae7dc9e7e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa709e1e209131cea4c470.md>)

[Implementation](<https://github.com/google-research/text-to-text-transfer-transformer>)

Recorded evaluation configuration: {"Chunk Size":"700","Latency (ms)":"108 ± 11"}. This record does not assert independent reproduction.

### Fixed

Precision: 72. Result ID: 6a13f755fbc373592c0ab647.

[Reporting paper](<https://www.sota2.com/research/paper/6a13d77e6f71c5ae7dc9e7e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa709e1e209131cea4c470.md>)

[Implementation](<https://github.com/google-research/text-to-text-transfer-transformer>)

Recorded evaluation configuration: {"Chunk Size":"300","Latency (ms)":"115 ± 12"}. This record does not assert independent reproduction.

### Fixed

Precision: 71. Result ID: 6a13f755fbc373592c0ab64a.

[Reporting paper](<https://www.sota2.com/research/paper/6a13d77e6f71c5ae7dc9e7e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa709e1e209131cea4c470.md>)

[Implementation](<https://github.com/google-research/text-to-text-transfer-transformer>)

Recorded evaluation configuration: {"Chunk Size":"1000","Latency (ms)":"105 ± 10"}. This record does not assert independent reproduction.

### Fixed

Precision: 70. Result ID: 6a13f755fbc373592c0ab646.

[Reporting paper](<https://www.sota2.com/research/paper/6a13d77e6f71c5ae7dc9e7e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa709e1e209131cea4c470.md>)

[Implementation](<https://github.com/google-research/text-to-text-transfer-transformer>)

Recorded evaluation configuration: {"Chunk Size":"150","Latency (ms)":"120 ± 15"}. This record does not assert independent reproduction.

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