Multimodal Weekly 89: Search for Reasoning Tasks and Reasoning via Continuous Concept Space
In the 89th session of Multimodal Weekly, we have two exciting presentations from researchers working in reasoning models:
✅ Rulin Shao and Rui Qiao will present their work called ReasonIR - the first retriever specifically trained for general reasoning tasks, achieving the state-of-the-art retrieval performance on BRIGHT (reasoning-intensive retrieval). When employed for retrieval-augmented generation (RAG), ReasonIR-8B also brings substantial gains on MMLU and GPQA.
ReasonIR Paper: https://arxiv.org/abs/2504.20595
ReasonIR Codebase: https://github.com/facebookresearch/ReasonIR?tab=readme-ov-file
✅ Zhen Zhang will present his work called "Soft Thinking" - a training-free method that emulates human-like "soft" reasoning by generating soft, abstract concept tokens in a continuous concept space. In essence, each generated concept token encapsulates multiple meanings from related discrete tokens, implicitly exploring various reasoning paths to converge effectively toward the correct answer.
Soft Thinking Paper: https://arxiv.org/abs/2505.15778
Soft Thinking Codebase: https://github.com/eric-ai-lab/Soft-Thinking
Join the Multimodal Minds community to connect with the speakers!
Multimodal Weekly is organized by Twelve Labs, a startup building multimodal foundation models for video understanding. Learn more about Twelve Labs here: https://twelvelabs.io/