Upload folder using huggingface_hub

.gitattributes

@@ -35,3 +35,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 checkpoint-1524/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 tokenizer.json filter=lfs diff=lfs merge=lfs -text
+train.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,62 +1,22 @@
 ---
-base_model: Qwen/Qwen2.5-Math-7B-Instruct
-library_name: peft
-model_name: verifier_v_7b
-tags:
-- base_model:adapter:Qwen/Qwen2.5-Math-7B-Instruct
-- lora
-- sft
-- transformers
-- trl
-licence: license
-pipeline_tag: text-generation
+license: mit
+task_categories: [text-generation]
+tags: [math, process-verification, ProcessBench, PRM800K, mathcompose]
 ---
 
-# Model Card for verifier_v_7b
+# mathcompose — Model V (process verifier) SFT data
 
-This model is a fine-tuned version of [Qwen/Qwen2.5-Math-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Math-7B-Instruct).
-It has been trained using [TRL](https://github.com/huggingface/trl).
+First-error-localization critiques for training a small generative math verifier.
+Each row: a problem + a step-indexed candidate solution (`prompt`) and a
+paragraph-by-paragraph critique ending in `\boxed{{k}}` (`completion`), where `k`
+is the 0-based index of the first erroneous step (`-1` = all correct).
 
-## Quick start
+- **Labels** are ground truth from **PRM800K** (OpenAI, MIT).
+- **Genuine-detection recipe:** **claude-opus-4-8** critiques each solution *blind*
+  (never shown the answer); a critique is kept only if its predicted index matches
+  the PRM800K gold (rejection sampling, ~77% keep-rate). So the critiques contain
+  real error-finding reasoning, not rationalization.
+- **Deduped** against ProcessBench + MATH-500 (no eval contamination).
+- Naturally ~50/50 erroneous/all-correct (no rebalancing needed).
 
-```python
-from transformers import pipeline
-
-question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
-generator = pipeline("text-generation", model="None", device="cuda")
-output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
-print(output["generated_text"])
-```
-
-## Training procedure
-
- 
-
-
-
-This model was trained with SFT.
-
-### Framework versions
-
-- PEFT 0.19.1
-- TRL: 1.8.0
-- Transformers: 5.13.1
-- Pytorch: 2.11.0+cu128
-- Datasets: 5.0.0
-- Tokenizers: 0.22.2
-
-## Citations
-
-
-
-Cite TRL as:
-    
-```bibtex
-@software{vonwerra2020trl,
-  title   = {{TRL: Transformers Reinforcement Learning}},
-  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
-  license = {Apache-2.0},
-  url     = {https://github.com/huggingface/trl},
-  year    = {2020}
-}
-```
+Built by the mathcompose harness. Eval target: ProcessBench first-error F1.

stats.json

@@ -0,0 +1,9 @@
+{
+  "n_total": 8554,
+  "n_train": 8127,
+  "n_val": 427,
+  "n_erroneous": 4316,
+  "n_all_correct": 4238,
+  "teacher": "promptlens",
+  "deduped": true
+}

train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5daa5a76a515a874eab6db9f445083bd2d9b704b4858aef2f7a14c646065a797
+size 20155092