Research map / Vision-language and multimodal / Multimodal LLMs
Transformation Driven Visual Reasoning
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a single-stream encoder rather than a two-stream encoder | TranceNet on TRANCE Basic · TRANCE Basic | Acc: 0.9571 -> 0.9810 | Helped |
| Add correctness-reward training with REINFORCE (ablation) | ResNet⊕ on TRANCE Event · TRANCE Event | Acc: 0.5785 -> 0.5978 | Helped |
| Add distance-reward training with REINFORCE (ablation) | ResNet⊕ on TRANCE Event · TRANCE Event | Acc: 0.5785 -> 0.5938 | Helped |
| Add both correctness and distance rewards with REINFORCE (ablation) | ResNet⊕ on TRANCE Event · TRANCE Event | Acc: 0.5785 -> 0.6001 | Helped |
| Use DUDA-style feature interaction instead of ResNet⊕ | TranceNet on TRANCE View · TRANCE View | Acc: 0.5305 -> 0.4590 | Hurt |
| Introduce variant final-camera views | deep models evaluated from Event to View · TRANCE View | Acc: 0.5785 -> 0.5305 | Hurt |
| Increase transformation sequence length | CNN− on TRANCE Event · TRANCE Event | LAcc: 95%, 56%, 23%, and 10% | Hurt |
| Predict all atomic transformations without attending to their order constraints | deep models on order-sensitive TRANCE Event samples · TRANCE Event, 7.8% order-sensitive samples | Acc: 0.3862 -> 0.3205 | Mixed |
| Use more training samples | ResNet⊕ on TRANCE Basic, Event, and View · TRANCE Basic, Event, and View | performance | Helped |
Closest papers by meaning
- CodeDance: A Dynamic Tool-integrated MLLM for Executable Visual Reasoning (CVPR 2026 · similarity 0.88)
- Multi-step Visual Reasoning with Visual Tokens Scaling and Verification (NeurIPS 2025 · similarity 0.88)
- STAR-R1: Multi-View Spatial TrAnsformation Reasoning by Reinforcing Multimodal LLMs (CVPR 2026 · similarity 0.87)
- Beyond the Doors of Perception: Vision Transformers Represent Relations Between Objects (NeurIPS 2024 · similarity 0.87)
- CogCoM: A Visual Language Model with Chain-of-Manipulations Reasoning (ICLR 2025 · similarity 0.87)
- CLiViS: Unleashing Cognitive Map through Linguistic-Visual Synergy for Embodied Visual Reasoning (CVPR 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.