technique

Reasoning Enhancement

techniqueactiveprovisionalreasoning-enhancement-c08a243c·1 events·first seen 22d ago

Aliases: Reasoning Enhancement

Co-occurring entities

Qwen3-4B ETCHR Reasoning Imitation Gemini 3.1 Flash Live Pass@1 Kimi K2.5

More like this (12)

Long-context Reasoning Benchmarks Learning to Reason by Analogy via Retrieval-Augmented Reinforcement Fine-Tuning Reasoning Language Models Reasoning in Memory (RiM)Reasoning Imitation MedCaseReasoning Predicting Future Behaviors in Reasoning Models Enables Better Steering Does Reasoning Preserve Alignment? On the Trustworthiness of Large Reasoning Models Agentic Chain-of-Thought Steering for Efficient and Controllable LLM Reasoning Failed Reasoning Traces Tell You What Is Fixable (But Not by Reading Them)Chain-of-Thought Reasoning Chain-of-Thought Fine-Tuning

Recent events (1)

6arXiv · cs.AI·22d ago·source ↗

ETCHR: Decoupled Image Editing for Visual Chain-of-Thought Reasoning in MLLMs

ETCHR introduces a question-conditioned, reasoning-aware image editing model that decouples visual transformation from downstream understanding in multimodal LLMs. It addresses two identified gaps—language-side (mapping abstract questions to visual edits) and generation-side (edit quality degrading with reasoning depth)—via a two-stage training recipe combining supervised fine-tuning on edit trajectories and VLM-derived reward signals. Because the editor is decoupled, it plugs into arbitrary MLLMs without retraining, yielding Pass@1 gains of roughly +4.6 to +5.5 points across five task families when paired with Qwen3-VL-8B, Gemini-3.1-Flash-Lite, and Kimi K2.5. The work advances the 'think with images' paradigm beyond fixed toolkits and unified multimodal approaches.

Agent and Tool Ecosystem Alignment and RLHF Reasoning Enhancement Qwen3-4B ETCHR +5 more