# Driving Question Answering on DRIVEXQA 1.0 (test) Source: [Driving Question Answering on DRIVEXQA 1.0 (test)]() Collection index: [llms.txt]() Data updated: 2026-03-13T07:48:15.055Z Evaluation of model performance on the test split of the DRIVEXQA 1.0 dataset for Driving Question Answering, reporting scores across multiple sub-categories (e.g., CL, FG, NT) based on both GPT-4o-mini evaluation and human assessment. 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 | Driving Question Answering | | Dataset (including any stated split) | DRIVEXQA 1.0 (test) | | Metric | CL Score (GPT-4o-mini) | | 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 - [CL Score (GPT-4o-mini)](): Higher is better; unit: PERCENT. - [CL Score (Human)](): Higher is better; unit: PERCENT. - [FG Score (GPT-4o-mini)](): Higher is better; unit: PERCENT. - [FG Score (Human)](): Higher is better; unit: PERCENT. - [NT Score (GPT-4o-mini)](): Higher is better; unit: PERCENT. - [NT Score (Human)](): Higher is better; unit: PERCENT. - [RN Score (GPT-4o-mini)](): Higher is better; unit: PERCENT. - [RN Score (Human)](): Higher is better; unit: PERCENT. - [SN Score (GPT-4o-mini)](): Higher is better; unit: PERCENT. - [SN Score (Human)](): Higher is better; unit: PERCENT. - [MB Score (GPT-4o-mini)](): Higher is better; unit: PERCENT. - [MB Score (Human)](): Higher is better; unit: PERCENT. - [OE Score (GPT-4o-mini)](): Higher is better; unit: PERCENT. - [OE Score (Human)](): Higher is better; unit: PERCENT. - [UE Score (GPT-4o-mini)](): Higher is better; unit: PERCENT. - [UE Score (Human)](): Higher is better; unit: PERCENT. - [LJ Score (GPT-4o-mini)](): Higher is better; unit: PERCENT. - [LJ Score (Human)](): Higher is better; unit: PERCENT. - [EL Score (GPT-4o-mini)](): Higher is better; unit: PERCENT. - [EL Score (Human)](): Higher is better; unit: PERCENT. - [Average Score (GPT-4o-mini)](): Higher is better; unit: PERCENT. - [Average Score (Human)](): Higher is better; unit: PERCENT. ## Recorded results | Method | CL Score (GPT-4o-mini) | Result date | Report paper | | --- | --- | --- | --- | | prepend | 61.2 | Not recorded | 69b399e70636e5bf4afbc8b9 | | MVX-LLM | 59.6 | Not recorded | 69b399e70636e5bf4afbc8b9 | | QAttn(Spectral) | 56.5 | Not recorded | 69b399e70636e5bf4afbc8b9 | | GAP | 51.2 | Not recorded | 69b399e70636e5bf4afbc8b9 | | GAP + Honeybee | 48.7 | Not recorded | 69b399e70636e5bf4afbc8b9 | | Self-query | 40.9 | Not recorded | 69b399e70636e5bf4afbc8b9 | | QAttn(DepthGate) | 38.3 | Not recorded | 69b399e70636e5bf4afbc8b9 | | GAP + Pargo | 36.5 | Not recorded | 69b399e70636e5bf4afbc8b9 | ### prepend CL Score (GPT-4o-mini): 61.2. Result ID: 69b3bf6a47f34ce9249d7924. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### MVX-LLM CL Score (GPT-4o-mini): 59.6. Result ID: 69b3bf6a47f34ce9249d792b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Projector Architectures"}. This record does not assert independent reproduction. ### QAttn(Spectral) CL Score (GPT-4o-mini): 56.5. Result ID: 69b3bf6a47f34ce9249d7927. [Reporting paper]() Recorded evaluation configuration: {"Category":"Token Fusion"}. This record does not assert independent reproduction. ### GAP CL Score (GPT-4o-mini): 51.2. Result ID: 69b3bf6a47f34ce9249d7926. [Reporting paper]() Recorded evaluation configuration: {"Category":"Token Fusion"}. This record does not assert independent reproduction. ### GAP + Honeybee CL Score (GPT-4o-mini): 48.7. Result ID: 69b3bf6a47f34ce9249d7929. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"Projector Architectures"}. This record does not assert independent reproduction. ### Self-query CL Score (GPT-4o-mini): 40.9. Result ID: 69b3bf6a47f34ce9249d7925. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"Token Fusion"}. This record does not assert independent reproduction. ### QAttn(DepthGate) CL Score (GPT-4o-mini): 38.3. Result ID: 69b3bf6a47f34ce9249d7928. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"Token Fusion"}. This record does not assert independent reproduction. ### GAP + Pargo CL Score (GPT-4o-mini): 36.5. Result ID: 69b3bf6a47f34ce9249d792a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Category":"Projector Architectures"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.