The paper introduces Scaffolding Minds, a method to improve multimodal reasoning by optimizing latent visual target representations. It addresses limitations in the supervised fine-tuning and reinforcement learning stages by using a dedicated scaffolding encoder and learning both mean and variance of the RL sampler, leading to significant performance gains on various benchmarks.
The chart will appear after repeat observations. The current metric comes from the source.
1 source votesReal observations only. History before source connection is not reconstructed.
This approach enhances multimodal reasoning by improving latent representation quality and exploration during reinforcement learning, leading to better performance on visual reasoning tasks.
The paper introduces VisionHOPE, a novel visual backbone that functions as a self-modifying learning system, allowing the model to co-evolve what it remembers and how it learns within an image. It uses five coupled memories and a stability-matched step-size control scheme to ensure stable learning dynamics, achieving competitive results on benchmark datasets like ImageNet-1K, COCO, and ADE20K.
3 days agoDaily PapersThe paper explores the linearity in Large Language Models (LLMs) by showing that combining inputs from different text streams leads to a superposition of next-token distributions. It suggests that this linearity is an inherent property of the Transformer architecture and can be restored through fine-tuning, allowing for generating two coherent continuations from one forward pass.
6 days agoDaily PapersThis paper introduces FuseReg, a method that replaces heuristic layer fusion in representation autoencoders (RAEs) with training over random subsets of encoder layers. The approach reduces the reconstruction-generation gap by improving robustness to layer fusion choices, achieving higher PSNR and lower generation FID scores without modifying the pretrained encoder.
5 days agoDaily PapersThe paper introduces Omni-IO Skills, a plug-and-play agent harness that enables existing agents to handle multiple modalities through hierarchical skills, standardized execution interfaces, and dependency-aware orchestration. It demonstrates significant improvements in input support and semantic quality scores when applied to GPT-5.6 Sol and Claude Sonnet 5 on the UniM-90 dataset.
5 days agoDaily PapersThe paper introduces GAGAR, a framework for quality-aware credit redistribution in code agent reinforcement learning (RL). It uses dynamic sampling and an SFT-trained agentic grader to rank test-passing trajectories, adjusting advantages to prioritize higher-quality implementations. The method was evaluated on large-scale industrial code agents with significant parameter counts.
4 days agoDaily PapersThis paper introduces SentZero, a sentence-centric vision-language pretraining framework designed for zero-shot multi-task analysis of chest X-rays. It enhances positive-pair diversity and mitigates false negatives through LLM-based sentence structuring and visual embedding modulation.
2 days ago