Modelpublic

varadsrivastava/lm-playschool-qwen3.5-2b-sft-dpo-vllm

sourceHugging Faceapache-2.0updated 1mo agoView on Hugging Face

0likes72downloads

Model Card

R2 (config variant for vLLM)

Part of a five-regime developmental sweep of post-training methods for dialogue-game competence (LM Playschool Challenge 2026, team DAIR).

Not a separate model. These are the same merged weights as `lm-playschool-qwen3.5-2b-sft-dpo` (R2), republished with a composite (vision-language) config.json so that vLLM will load them. At the time of our experiments, vLLM's Qwen3.5 integration expected the composite config while transformers writes a text-only one; neither could read the other's schema. Use this repo only if you need vLLM; use the R2 repo for transformers. Scores are those of R2.

All numbers are clemscore / statscore on the playpen validation split, measured in a single frozen environment (Python 3.11, clemcore pinned via playpen, clembench pinned requirements) with two upstream fixes applied: a division-by-zero guard in the privateshared Game Master and the punkt_tab NLTK resource for the IFEval scorer. Earlier revisions of this card reported numbers from an unpinned environment; see the paper for the environment-sensitivity analysis.

Checkpoint family (LM Playschool challenge, team DAIR)

Regime	Repo	clem	stat
R1 imitation (SFT)	`lm-playschool-qwen3.5-2b-sft`	55.61	43.87
R2 outcome contrast (DPO)	`lm-playschool-qwen3.5-2b-sft-dpo`	67.39	44.72
R3 self-imitation (SFT)	`lm-playschool-qwen3.5-2b-iter3`	61.06	44.01
R4 corrective feedback (DPO)	`lm-playschool-qwen3.5-2b-iter4`	67.64	44.31
R5 GRPO (control)	`lm-playschool-qwen3.5-2b-grpo-base-s42`	62.43	44.19
R5 GRPO + RND	`lm-playschool-qwen3.5-2b-grpo-rnd-s42`	67.44	43.53

Base model: Qwen3.5-2B (13.63 / 44.22 in the same environment). Paper: Raising a Small Language Model: From Imitation to Curiosity in Dialogue Games (LM Playschool Challenge 2026).