# Continual Visual Instruction Tuning on CVIT Single-type instruction Source: [Continual Visual Instruction Tuning on CVIT Single-type instruction]() Collection index: [llms.txt]() Data updated: 2026-05-15T06:26:11.421Z Evaluation of model performance on the CVIT Single-type instruction dataset, focusing on Continual Visual Instruction Tuning capabilities, measured across standard VQA benchmarks (ScienceQA, VQAv2, GQA) and specialized metrics (BWT, MIF). 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 | Continual Visual Instruction Tuning | | Dataset (including any stated split) | CVIT Single-type instruction | | Metric | AP (Average Precision) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 16 of 16 | | 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 - [AP (Average Precision)](): Higher is better; unit: PERCENT. - [ScienceQA Accuracy](): Higher is better; unit: PERCENT. - [TextVQA Accuracy](): Higher is better; unit: PERCENT. - [Flickr30k Score](): Higher is better; unit: SCALAR. - [ImageNet Accuracy](): Higher is better; unit: PERCENT. - [GQA Accuracy](): Higher is better; unit: PERCENT. - [VQAv2 Accuracy](): Higher is better; unit: PERCENT. - [MAP (Mean Average Precision)](): Higher is better; unit: PERCENT. - [BWT (Backward Transfer)](): Higher is better; unit: SCALAR. - [MIF](): Higher is better; unit: PERCENT. ## Recorded results | Method | AP (Average Precision) | Result date | Report paper | | --- | --- | --- | --- | | DirLoRA\* | 88.13 | Not recorded | 68fa7a881e209131cea83bd1 | | LiLoRA | 84.88 | Not recorded | 68fa7a881e209131cea83bd1 | | SMoLoRA | 83.44 | Not recorded | 68fa7a881e209131cea83bd1 | | DirLoRA | 67.55 | Not recorded | 68fa7a881e209131cea83bd1 | | Eproj | 65.29 | Not recorded | 68fa7a881e209131cea83bd1 | | LiLoRA | 64.63 | Not recorded | 68fa7a881e209131cea83bd1 | | Eproj | 58.87 | Not recorded | 68fa7a881e209131cea83bd1 | | Replay | 49.7 | Not recorded | 68fa7a881e209131cea83bd1 | | EWC+TIR | 47.19 | Not recorded | 68fa7a881e209131cea83bd1 | | EWC | 46.26 | Not recorded | 68fa7a881e209131cea83bd1 | | SeqLoRA | 46.21 | Not recorded | 68fa7a881e209131cea83bd1 | | C-LoRA | 45.78 | Not recorded | 68fa7a881e209131cea83bd1 | | SeqLoRA | 45.41 | Not recorded | 68fa7a881e209131cea83bd1 | | MoeLoRA | 45.27 | Not recorded | 68fa7a881e209131cea83bd1 | | DoRA | 44.07 | Not recorded | 68fa7a881e209131cea83bd1 | | Zero-shot | 22.3 | Not recorded | 68fa7a881e209131cea83bd1 | ### DirLoRA\* AP (Average Precision): 88.13. Result ID: 6a06bb9947f34ce924eda28f. [Reporting paper]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. ### LiLoRA AP (Average Precision): 84.88. Result ID: 6a06bb9947f34ce924eda299. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. ### SMoLoRA AP (Average Precision): 83.44. Result ID: 6a06bb9947f34ce924eda298. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. ### DirLoRA AP (Average Precision): 67.55. Result ID: 6a06bb9947f34ce924eda2a6. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen2-VL-2B","Instruction Setting":"Single-type instruction setting"}. This record does not assert independent reproduction. ### Eproj AP (Average Precision): 65.29. Result ID: 6a06bb9947f34ce924eda297. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. ### LiLoRA AP (Average Precision): 64.63. Result ID: 6a06bb9947f34ce924eda2a9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2-VL-2B","Instruction Setting":"Single-type instruction setting"}. This record does not assert independent reproduction. ### Eproj AP (Average Precision): 58.87. Result ID: 6a06bb9947f34ce924eda2a8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2-VL-2B","Instruction Setting":"Single-type instruction setting"}. This record does not assert independent reproduction. ### Replay AP (Average Precision): 49.7. Result ID: 6a06bb9947f34ce924eda294. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. ### EWC+TIR AP (Average Precision): 47.19. Result ID: 6a06bb9947f34ce924eda296. [Reporting paper]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. ### EWC AP (Average Precision): 46.26. Result ID: 6a06bb9947f34ce924eda295. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. ### SeqLoRA AP (Average Precision): 46.21. Result ID: 6a06bb9947f34ce924eda290. [Reporting paper]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. ### C-LoRA AP (Average Precision): 45.78. Result ID: 6a06bb9947f34ce924eda293. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. ### SeqLoRA AP (Average Precision): 45.41. Result ID: 6a06bb9947f34ce924eda2a7. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen2-VL-2B","Instruction Setting":"Single-type instruction setting"}. This record does not assert independent reproduction. ### MoeLoRA AP (Average Precision): 45.27. Result ID: 6a06bb9947f34ce924eda292. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. ### DoRA AP (Average Precision): 44.07. Result ID: 6a06bb9947f34ce924eda291. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. ### Zero-shot AP (Average Precision): 22.3. Result ID: 6a06bb9947f34ce924eda28e. [Reporting paper]() Recorded evaluation configuration: {"Instruction Setting":"Single-type"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.