SIMIT: Self-Improving Vision-Language Models via Imagination at Test-Time
cs.CV, cs.LG
Submitted: 2026-10-11
Updated: 2026-10-11
Comments: Project Page: https://monurcan.github.io/simit
Project page: https://monurcan.github.io/simit
License: http://creativecommons.org/licenses/by/4.0/
The gist: We introduce SIMIT, a test-time self-improvement framework in which a vision-language model creates its own query-specific training data before answering an unlabeled test query.
Terminology
Abstract
We introduce SIMIT, a test-time self-improvement framework in which a vision-language model creates its own query-specific training data before answering an unlabeled test query. Our goal is to improve performance on a given target test set without annotations or external models. To pursue this goal, some test-time methods improve answers through longer reasoning or repeated sampling but leave the model unchanged. Others update the model using rewards based on majority agreement among multiple candidate answers or the model's own judgments of correctness, but such rewards can reinforce existing mistakes. SIMIT instead imagines similar problems whose solutions it controls: a model need not correctly solve the target query to construct useful practice examples. It synthesizes diverse (question, answer, image) triplets by specifying each answer before realizing its image. To support diverse multimodal tasks, SIMIT combines native image generation with an extensible skill library for structured visuals, including documents, charts, and diagrams. Within this pipeline, multi-stage verification checks that images support their assigned answers, while difficulty filtering retains informative examples. Adaptive budgeting further targets this practice by allocating more samples to queries the model is less confident in answering. The resulting data can improve the model in-context (SIMIT-ICL) or in-weight (SIMIT-FT). Across 17 diverse benchmarks, SIMIT outperforms existing self-improvement methods. Its best configuration achieves a +7.20% mean relative gain over BAGEL-7B, while training-free SIMIT-ICL achieves +6.98%.
Sources
- Large Language Monkeys: Scaling Inference Compute with Repeated Sampling
- Training Verifiers to Solve Math Word Problems
- Emerging Properties in Unified Multimodal Pretraining
- SenseNova-U1: Unifying Multimodal Understanding and Generation with NEO-unify Architecture
- Lance: Unified Multimodal Modeling by Multi-Task Synergy
- AVIS: Adaptive Test-Time Scaling for Vision-Language Models
- M*: A Modular, Extensible, Serving System for Multimodal Models
- Self-Generated In-Context Learning: Leveraging Auto-regressive Language Models as a Demonstration Generator
- Zebra-CoT: A Dataset for Interleaved Vision Language Reasoning
- SEED-Bench: Benchmarking Multimodal LLMs with Generative Comprehension
- Qwen3-VL-Embedding and Qwen3-VL-Reranker: A Unified Framework for State-of-the-Art Multimodal Retrieval and Ranking
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence
- Test-time Scaling of LLMs: A Survey from A Subproblem Structure Perspective
- What Makes Good Agentic Data? An ACE Lens on Data Generation for LLM Agents
- A Survey on Test-Time Scaling in Large Language Models: What, How, Where, and How Well?
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs