Stanford and Together AI's HybridAgents: 10.4% gains, up to 3.7x cost savings via latent reasoning
StanfordUniversity · hf · 2026-10-15
Researchers from Stanford, Together AI, UIUC and Meta (including James Zou) introduce HybridAgents, agents that dynamically navigate between latent reasoning and text-based interaction.
- A lightweight HybridLink enables seamless text↔latent transitions, while continuous working memory preserves lossless hybrid context across turns—keeping tool access while making latent thoughts inspectable.
- Post-training pairs HybridSFT (stabilizing tool-integrated hybrid generation) with HybridRL (learning when to think in latent space vs. act via text).
- Across 3 model scales and 11 benchmarks (math, science, medicine, search, agentic reasoning), HybridAgents improves performance by 10.4% vs. frontier text- and latent-based agents, with 35.5–54.7% token reduction, 1.8–2.6x faster inference, and 2.7–3.7x cost savings. Code and data are open-sourced.
More from coding & agent
- Alibaba's HICO-Bench shows 36% of highly-rated agent conversations fail verification — Tongyi-ConvAI · 2026-10-15
- ByteDance's Diff Diffusion predicts code edit paths: 2.78x faster inference, 39% lower latency — Xiaolong Xu · 2026-10-15
- Reddit debate: AI agent builders are forgetting basic software engineering — Dangerous_Bus4157 · 2026-10-15
- Gemini CLI PR fixes background Git update checks hijacking the TUI via hidden auth prompts — villahernandez-coder · 2026-10-15
- Player-1: RL turns VLM agents into GTA V players with zero game-specific training — yshan2u · 2026-10-15
- Agentic engineering vs AI engineering: one builds software with agents, the other builds agents as the product — Al_Grigor · 2026-10-15