Open-weight model
Qwen2.5-3B-Instruct-Sheldon-RLVR-math-v1-LoRA
by Sam EnMin Huang tbooy/Qwen2.5-3B-Instruct-Sheldon-RLVR-math-v1-LoRA
Qwen2.5-3B-Instruct-Sheldon-RLVR-math-v1-LoRA is an open-weight model from Sam EnMin Huang, released under other. Its published files total 2.2 GB.
Files: finaladapter/ (PEFT LoRA, step 200) and checkpoint-{25,...,200}/ adapters with trainer states (per-step reward, pass-rate and group statistics under rlvr/), plus rlvrargs.json. Apply to tbooy/Qwen2.5-3B-Instruct-Sheldon-RLAIF-grpo-v2.
Model Card
Files: finaladapter/ (PEFT LoRA, step 200) and checkpoint-{25,...,200}/ adapters with trainer states (per-step reward, pass-rate and group statistics under rlvr/), plus rlvrargs.json. Apply to tbooy/Qwen2.5-3B-Instruct-Sheldon-RLAIF-grpo-v2. Merged weights: tbooy/Qwen2.5-3B-Instruct-Sheldon-RLVR-math-v1. Stage-3 (RLVR) model of a Harvard CS 2881R project: reinforcement learning with a verifiable math reward on top of tbooy/Qwen2.5-3B-Instruct-Sheldon-RLAIF-grpo-v2 (SFT persona model -> RLAIF GRPO -> this), the persona being Dr. Sheldon Cooper. Why this stage exists: the persona SFT stage cost most of Qwen2.5-3B-Instruct's competition-math ability (MATH-500 greedy 68.0 -> 35.4), and the…
Excerpt from the card by Sam EnMin Huang, licensed other.
Identity and Version
- Repository
- tbooy/Qwen2.5-3B-Instruct-Sheldon-RLVR-math-v1-LoRA
- Publisher
- Sam EnMin Huang
- Task
- Not stated by the source
- Modality
- Other
- Library
- peft
- Parameters
- Not stated by the source
- Languages
- en
- Revision
- a5ed9df0da889c659c647082cf57ebe5e84f9376
- First published
- 2026-09-26
- Last updated
- 2026-09-26
Files and Weights
33 files, 2.2 GB in total. The weights are 9 files totalling 2.2 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| checkpoint-100/adapter_model.safetensors | Weights | 239.5 MB | 379e61210b8c |
| checkpoint-125/adapter_model.safetensors | Weights | 239.5 MB | 21cdd2515f5c |
| checkpoint-150/adapter_model.safetensors | Weights | 239.5 MB | edc1665ce72c |
| checkpoint-175/adapter_model.safetensors | Weights | 239.5 MB | a2eb2bc609ff |
| checkpoint-200/adapter_model.safetensors | Weights | 239.5 MB | 26386cbd5b24 |
| checkpoint-25/adapter_model.safetensors | Weights | 239.5 MB | 4f2d7cdbf2e1 |
| checkpoint-50/adapter_model.safetensors | Weights | 239.5 MB | 29f44943a6f6 |
| checkpoint-75/adapter_model.safetensors | Weights | 239.5 MB | 1b84fc0ca5ab |
| final_adapter/adapter_model.safetensors | Weights | 239.5 MB | 26386cbd5b24 |
| checkpoint-100/adapter_config.json | Configuration | 1.2 KB | — |
| checkpoint-100/trainer_state.json | Configuration | 172.0 KB | — |
| checkpoint-125/adapter_config.json | Configuration | 1.2 KB | — |
| checkpoint-125/trainer_state.json | Configuration | 215.2 KB | — |
| checkpoint-150/adapter_config.json | Configuration | 1.2 KB | — |
| checkpoint-150/trainer_state.json | Configuration | 258.5 KB | — |
| checkpoint-175/adapter_config.json | Configuration | 1.2 KB | — |
| checkpoint-175/trainer_state.json | Configuration | 302.0 KB | — |
| checkpoint-200/adapter_config.json | Configuration | 1.2 KB | — |
| checkpoint-200/trainer_state.json | Configuration | 346.4 KB | — |
| checkpoint-25/adapter_config.json | Configuration | 1.2 KB | — |
| checkpoint-25/trainer_state.json | Configuration | 43.8 KB | — |
| checkpoint-50/adapter_config.json | Configuration | 1.2 KB | — |
| checkpoint-50/trainer_state.json | Configuration | 86.5 KB | — |
| checkpoint-75/adapter_config.json | Configuration | 1.2 KB | — |
| checkpoint-75/trainer_state.json | Configuration | 129.1 KB | — |
| final_adapter/adapter_config.json | Configuration | 1.2 KB | — |
| rlvr_args.json | Configuration | 1.2 KB | — |
| README.md | Documentation | 3.7 KB | — |
| final_adapter/README.md | Documentation | 5.2 KB | — |
| final_adapter/chat_template.jinja | Other | 2.5 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| final_adapter/tokenizer.json | Tokenizer | 11.4 MB | 3fd169731d2c |
| final_adapter/tokenizer_config.json | Tokenizer | 693 B | — |
License and Download
- License
- other
- Access
- Open weights, no gate
- Download size
- 2.2 GB
Released by Sam EnMin Huang through its official repository on Hugging Face. Read the license.
Built From
- Adapter of tbooy/Qwen2.5-3B-Instruct-Sheldon-RLAIF-grpo-v2
- Derived from tbooy/Qwen2.5-3B-Instruct-Sheldon-RLAIF-grpo-v2
- Trained on (disclosed) HuggingFaceH4/MATH-500
- Trained on (disclosed) openai/gsm8k
- Trained on (disclosed) rasbt/math_full_minus_math500
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 2.2 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About Qwen2.5-3B-Instruct-Sheldon-RLVR-math-v1-LoRA
What license is Qwen2.5-3B-Instruct-Sheldon-RLVR-math-v1-LoRA released under?
other, as its publisher declares it. Read the license text before commercial use.