Call for Participation: LM Playschool (LMP 2026)
Improving Language Models through Learning from Dialogue Interaction Co-located with EMNLP 2026 — 24-29 October 2026, Budapest
Website: https://lm-playschool.github.io/
NEW! Starter Kit: https://github.com/lm-playpen/playpen
The LM Playschool Workshop (LMP 2026) invites submissions exploring the frontier of language agents that learn and adapt through situated interaction. Following our first announcement, we are excited to issue this second call and officially release the challenge Starter Kit.
📝 SUBMISSION TRACKS We welcome either long or short submissions for the following tracks: 1. Challenge track: Technical reports for the LM Playschool challenge (archival). 2. Paper-only track: Work-in-progress (archival or non-archival) or recently published papers (non-archival).
🏆 CHALLENGE TRACK (SHARED TASK) The shared task focuses on post-training LLMs to master communicative skills in unseen dialogue games while retaining original language capabilities. Participants are free to choose any base model; evaluation is based on improvement relative to that base model on an unseen test set.
Sign up here: https://forms.gle/fhpXPH5kZk4psPXp9
🛠️ THE STARTER KIT (NEW!): To support participants in the LM Playschool Challenge, we have released "Playpen," a sandbox designed to bridge the gap between interaction and training. Whether you are interested in Reinforcement Learning (RL), Supervised Fine-Tuning (SFT), or zero-shot evaluation, Playpen simplifies the workflow so you can focus on the science.
Key features include: * Comprehensive Evaluation: A single command to get your "clemscore" (interactive competence) and "statscore" (classic benchmarks). * Training Recipes: Example scripts for SFT and GRPO (Group Relative Policy Optimization) to help models learn from game-state success. * Resource Friendly: We encourage using the Qwen3.5 family (0.8B to 27B) to ensure participation is possible with modest compute.
🎯PAPER-ONLY TRACK: TOPICS OF INTEREST We welcome original research and work-in-progress on: * Architectures and training regimes for interactive agents. * Intrinsic rewards and learning signals (RL from game-state success). * Benchmarking via dialogue games. * Data efficiency and social interaction. * Social cognition and Theory of Mind in interactive systems. * Human-agent collaboration and coordination. * Embodied interactive agents. * Communicative and perceptual grounding.
📅 IMPORTANT DATES * Starter pack released! Available on Github: https://github.com/lm-playpen/playpen * ARR paper submission deadline: May 25, 2026 * Challenge submission deadline: July 10, 2026 * Direct paper submission deadline: July 17, 2026 * Notification of acceptance: August 20, 2026 * Camera ready due: September 20, 2026 * Workshop at EMNLP 2026: October 24-29, 2026 (Budapest)
For more information, visit our website: https://lm-playschool.github.io/
We look forward to your submissions!
The LMP 2026 Organizing Committee: Raffaella Bernardi, Raquel Fernández, Mario Giulianelli, Sherzod Hakimov, Alexander Koller, Dieu-Thu Le, Oliver Lemon, Davide Mazzaccara, Sabrina McCallum, David Schlangen, Alessandro Suglia.