# Multi-hop reasoning and question answering on GAIA

Source: [Multi-hop reasoning and question answering on GAIA](<https://www.sota2.com/research/sota/multi-hop-reasoning-and-question-answering-on-gaia>)

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

Data updated: 2026-02-26T05:45:39.683Z

Evaluation of model performance on multi-hop reasoning and question answering tasks using the GAIA dataset, measured by GAIA Level 1, Level 2, Level 3, and Average accuracy scores.

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 | Multi-hop reasoning and question answering |
| Dataset (including any stated split) | GAIA |
| Metric | GAIA L1 Acc |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 2 of 2 |
| 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

- [GAIA L1 Acc](<https://www.sota2.com/research/sota/multi-hop-reasoning-and-question-answering-on-gaia.md?metric=GAIA%20L1%20Acc>): Higher is better; unit: PERCENT.

- [GAIA L2 Acc](<https://www.sota2.com/research/sota/multi-hop-reasoning-and-question-answering-on-gaia.md?metric=GAIA%20L2%20Acc>): Higher is better; unit: PERCENT.

- [GAIA L3 Acc](<https://www.sota2.com/research/sota/multi-hop-reasoning-and-question-answering-on-gaia.md?metric=GAIA%20L3%20Acc>): Higher is better; unit: PERCENT.

- [GAIA Avg Acc](<https://www.sota2.com/research/sota/multi-hop-reasoning-and-question-answering-on-gaia.md?metric=GAIA%20Avg%20Acc>): Higher is better; unit: PERCENT.

## Recorded results

| Method | GAIA L1 Acc | Result date | Report paper |
| --- | --- | --- | --- |
| AgentConductor | 72 | Not recorded | 6997ea6809c33900fcd7649f |
| Chain-of-Agents | 69.2 | Not recorded | 6997ea6809c33900fcd7649f |

### AgentConductor

GAIA L1 Acc: 72. Result ID: 699b05eb5698665ec49848f0.

[Reporting paper](<https://www.sota2.com/research/paper/6997ea6809c33900fcd7649f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6997ea6809c33900fcd7649f.md>)

Recorded evaluation configuration: {"Backbone":"3B"}. This record does not assert independent reproduction.

### Chain-of-Agents

GAIA L1 Acc: 69.2. Result ID: 699b05eb5698665ec49848ef.

[Reporting paper](<https://www.sota2.com/research/paper/6997ea6809c33900fcd7649f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7abc1e209131cea84e86.md>)

Recorded evaluation configuration: {"Backbone":"7B"}. This record does not assert independent reproduction.

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