Microsoft open-sources UserRepGen-4B, a user-representation generator for LLM user simulators
jacek2023 · reddit · 2026-10-15
Microsoft released UserRepGen-4B, a Qwen3.5-4B fine-tune that produces a structured natural-language user representation (task description + behavior description) from multi-turn user-assistant conversations, intended to condition a separate LLM-based user simulator.
- Unlike Microsoft's earlier UserLM-8b and UserLM-2.0, it generates representations rather than simulated user utterances; it doesn't execute tasks or act as the simulator itself.
- Trained with online supervised fine-tuning via hindsight distillation, using feedback from comparing simulated and original user behavior; at inference it works directly from the source conversation with no simulation rollouts.
- The architecture chart highlights use cases like simulating users in unseen scenarios, diverse user populations, and distinguishing users across conversations.
Open-sourced on Hugging Face.
Related event: Microsoft Open-Sources UserRepGen-4B User Profiling Model(2 posts)→
More from Models
- Beam's Ten-Benchmark Breakdown: Doubles Peers on CritPt, Near-Bottom on Terminal-Bench — ArtificialAnlys · 2026-10-16
- Beam Scores 28 on Coding Agent Index, 40 Behind Leaders but Among Most Token-Frugal — ArtificialAnlys · 2026-10-16
- Beam Token Breakdown: Only 26% of Its 26k Output Tokens per Task Are Answers — ArtificialAnlys · 2026-10-16
- Reflection's First Model Beam Scores 30, Ranks #2 Among US Open Weights Models — ArtificialAnlys · 2026-10-16
- Why are sub-15B open models so rare now? Labs seem to have settled on 24-27B minimum — Hot_Example_4456 · 2026-10-16
- User spots model openly generating "ideal responses" to train other models — liminal_bardo · 2026-10-16