| Paper | Code | Accuracy | ModelName | ReleaseDate |
|---|---|---|---|---|
| Priming Base Models for Agentic Post-Training: An Open Midtraining Recipe | 62.82 | NVIDIA Nemotron 3.5 Super VL 120B-A12B BF16 | 2026-10-15 | |
| InSight-doc: Agentic Visual Perception for Long-Document Understanding | ✓ Link | 57.9 | InSight-doc-8B (SFT+RL, r=0.7) (extra data) | 2026-08-11 |
| Nemotron 3 Nano Omni: Efficient and Open Multimodal Intelligence | ✓ Link | 57.5 | Nemotron 3 Nano Omni (30B-A3B, reasoning on) | 2026-04-27 |
| DocDancer: Towards Agentic Document-Grounded Information Seeking | 57.0 | DocDancer (GPT-5.2) | 2026-01-08 | |
| DocPrune: Efficient Document Question Answering via Background, Question, and Comprehension-aware Token Pruning | 32.2 | DocPrune (Qwen2.5-VL 7B) | 2026-04-24 | |
| NVIDIA Nemotron Nano V2 VL | ✓ Link | 32.1 | Nemotron Nano V2 VL (Reasoning off) (extra data) | 2025-11-06 |
| LoMo: Local Modality Substitution for Deeper Vision-Language Fusion | ✓ Link | 18.06 | LoMo (LLaVA-OneVision-1.5-8B) | 2026-05-28 |