Topic
Qwen3 8b
3 episodes
-
JoyAI VL Interaction: Real Time Vision Language Interaction Intelligence
JoyAI-VL-Interaction is an 8B vision-language model that flips the script on AI assistants: instead of waiting to be asked, it continuously watches a video stream and decides moment-to-moment whether to respond, stay silent, or delegate complex tasks to a background model. The system is fully deployable, open-sourced (weights, training recipe, data, and complete stack), and beats Doubao and Gemini's in-app video assistants on quality and timing in real-world scenarios.
-
LatentSkill: From In Context Textual Skills to In Weight Latent Skills for LLM Agents
Exploring LatentSkill, a framework that turns textual agent skills into weight-space LoRA adapters, cutting prompt overhead while keeping modularity and composability. Cody digs into the hypernetwork design and trade-offs; Justy asks what shipping this looks like and who’d actually adopt it.
-
How to Fine Tune a Reasoning Model? A Teacher–Student Cooperation Framework to Synthesize Student Consistent SFT Data
Episode 300 of Exploring Next digs into TESSY, a teacher-student data synthesis method for fine-tuning reasoning models without wrecking the smaller model’s existing style. The hosts unpack why direct teacher-generated supervised fine-tuning can actually make reasoning models worse, how TESSY alternates teacher-generated capability tokens with student-generated style tokens, and why that matters for anyone trying to ship smaller, cheaper reasoning systems for coding and other structured tasks.