--- license: apache-2.0 base_model: OpenGVLab/InternVL3-2B library_name: peft tags: - lora - polyvalent - internvl3 - logical-reasoning - deduction - reasoning pipeline_tag: text-generation --- # Polyvalent Logical Reasoning Specialist Adapter (v3_sft) PEFT LoRA adapter enhancing formal deduction, propositional logic, and puzzle-solving on the **InternVL3-2B** base model. ## Model Description - **Adapter Name:** `v3_sft` (Reasoning Specialist) - **Base Model:** `OpenGVLab/InternVL3-2B` (pinned at revision `899155015275a9b7338c7f4677e19c784e0e5a21`) - **LoRA Configuration:** Rank $r=64$, $\alpha=128$, dropout $0.05$ on `q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`. - **Base Weights:** Apache-2.0 ## Training Mixture Attributions Trained on a curated multi-source reasoning mixture: - **Dolly-15k:** Databricks (CC BY-SA 3.0) - **ProofWriter:** Allen Institute for AI (CC BY 4.0) - **FOLIO:** First-Order Logic dataset (CC BY-SA 4.0) - **Self-Generated Puzzles:** Synthetic propositional logic problems. - *Notice:* Adapter weights are released under Apache-2.0. Any downstream derivative works incorporating the training mixture data fall under CC BY-SA 4.0. ## Benchmark Performance Evaluated against rigorous deduction benchmarks: - **LogicBench:** Multi-step symbolic constraint satisfaction - **LogicNLI:** Natural language inference with formal reasoning - **Multi-LogiEval:** Multi-hop deductive reasoning suites ## How to Load ```python import torch from transformers import AutoModel, AutoTokenizer from peft import PeftModel BASE_MODEL = "OpenGVLab/InternVL3-2B" BASE_REVISION = "899155015275a9b7338c7f4677e19c784e0e5a21" tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL, revision=BASE_REVISION, trust_remote_code=True) base_model = AutoModel.from_pretrained( BASE_MODEL, revision=BASE_REVISION, torch_dtype=torch.bfloat16, trust_remote_code=True ).eval() model = PeftModel.from_pretrained(base_model, "Pushkarrrr15151/PolyVL-logic-lora").eval() ``` ## Citation ```bibtex @article{polyvalent2026, title={Polyvalent: Single-Base Multi-Specialist Architecture with Dynamic Adapter Residency}, author={Polyvalent Research Team}, year={2026} } ```