Add Qwen3-8B-CalibSFT-RLCR

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,98 @@
+---
+license: apache-2.0
+base_model: SUSTech/Qwen3-8B-CalibSFT
+base_model_relation: finetune
+library_name: transformers
+pipeline_tag: text-generation
+language:
+  - en
+datasets:
+  - SUSTech/CalibSFT-DeepScaleR
+tags:
+  - reasoning
+  - math
+  - confidence-calibration
+  - verbalized-confidence
+---
+
+# Qwen3-8B-CalibSFT-RLCR
+
+Qwen3-8B trained with **CalibSFT followed by RLCR**, from the paper [On the Pitfalls of Verbalized Confidence Priors for Calibrating Large Reasoning Models](https://arxiv.org/abs/XXXX.XXXXX). The model solves a problem, then states its answer and a verbalized confidence between 0 and 1. It is initialized from [SUSTech/Qwen3-8B-CalibSFT](https://huggingface.co/SUSTech/Qwen3-8B-CalibSFT).
+
+- **Code:** https://github.com/ml-stat-Sustech/verbalized-confidence-training
+- **Training data:** [SUSTech/CalibSFT-DeepScaleR](https://huggingface.co/datasets/SUSTech/CalibSFT-DeepScaleR)
+
+## Usage
+
+The model expects the system prompt and user format used during training:
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model_id = "SUSTech/Qwen3-8B-CalibSFT-RLCR"
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto", device_map="auto")
+
+SYSTEM_PROMPT = (
+    "A conversation between User and Assistant. The user asks a question, and the Assistant solves it. "
+    "The assistant first thinks about the reasoning process in the mind and analyzes its confidence about "
+    "the solution and then provides the user with the final answer as well as its confidence level. "
+    "The confidence level indicates how certain the Assistant is about its answer, expressed as a decimal "
+    "between 0 and 1 with exactly two decimal places, enclosed within <confidence> </confidence> tags. "
+    "The response must strictly follow this format: <think> reasoning process here </think> "
+    "<answer> final short answer only </answer> <confidence> 0.xx </confidence>. "
+    "The <answer> tag must contain only the final answer string needed for exact-match evaluation, "
+    "not a full sentence, explanation, or reasoning."
+)
+question = "What is the smallest positive integer n such that n^2 + n is divisible by 12?"
+messages = [
+    {"role": "system", "content": SYSTEM_PROMPT},
+    {"role": "user", "content": f"\n\nPROBLEM: {question}\n\n"},
+]
+
+text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+inputs = tokenizer(text, return_tensors="pt").to(model.device)
+outputs = model.generate(**inputs, max_new_tokens=16384, do_sample=True, temperature=0.6, top_p=0.95, top_k=20)
+print(tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True))
+# Output format: <think> ... </think> <answer> ... </answer> <confidence> 0.xx </confidence>
+```
+
+## Training
+
+The model starts from [SUSTech/Qwen3-8B-CalibSFT](https://huggingface.co/SUSTech/Qwen3-8B-CalibSFT) and is trained with RLCR on DeepScaleR. RLCR rewards answer correctness and format together with a Brier calibration reward on the stated confidence. Training follows the DAPO recipe with dynamic sampling, and questions that the base model already solves in every sampled response are removed beforehand.
+
+| Hyperparameter | Value |
+| --- | --- |
+| Responses per question | 8 |
+| Questions per batch | 64 |
+| Mini-batch size | 32 |
+| Steps | 160 |
+| Learning rate | 2e-6, cosine decay to 1e-6 after 10 warm-up steps |
+| Clip ratio (low / high) | 0.2 / 0.28 |
+| KL coefficient | 0 |
+| Max response length | 8,192 tokens |
+| Seed | 45 |
+
+## Evaluation
+
+Temperature 0.6, 4 responses per question (32 on AIME). Metrics are percentages averaged over 8 in-distribution math benchmarks (DeepScaleR-Eval, MATH-500, MinervaMath, OlympiadBench, GSM8K, AIME 2024–2026) and 8 out-of-distribution benchmarks (HotpotQA, TriviaQA, DROP, MuSiQue, LiveBenchReasoning, NQOpen, PopQA, WebQuestions).
+
+| Model | ID Pass@1 ↑ | ID AUROC ↑ | ID Brier ↓ | ID ECE ↓ | OOD Pass@1 ↑ | OOD AUROC ↑ | OOD Brier ↓ | OOD ECE ↓ |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| Qwen3-8B | 59.32 | 71.90 | 33.20 | 34.36 | 45.36 | 65.71 | 45.15 | 46.94 |
+| Qwen3-8B-RLCR | 65.33 | 82.36 | 14.43 | 14.46 | 43.96 | 69.38 | 31.14 | 31.17 |
+| Qwen3-8B-CalibSFT | 58.51 | 79.93 | 17.30 | 14.23 | 46.37 | 61.20 | 29.15 | 25.49 |
+| **Qwen3-8B-CalibSFT-RLCR** | 67.20 | 86.32 | 12.77 | 8.59 | 45.71 | 71.31 | 23.24 | 18.20 |
+
+Qwen3-8B and CalibSFT were evaluated with up to 6,144 generated tokens and the RLCR models with up to 16,384.
+
+## Citation
+
+```bibtex
+@article{wang2026pitfalls,
+  title={On the Pitfalls of Verbalized Confidence Priors for Calibrating Large Reasoning Models},
+  author={Wang, Shuoyuan and Luo, Beier and Zeng, Hao and Yu, Chengyao and Zhang, Songxin and Xie, Zejian and Jing, Bingyi and Wei, Hongxin},
+  journal={arXiv preprint arXiv:XXXX.XXXXX},
+  year={2026}
+}
+```

chat_template.jinja

@@ -0,0 +1,89 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}

config.json

@@ -0,0 +1,72 @@
+{
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "attn_implementation": "flash_attention_2",
+  "bos_token_id": null,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 12288,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 40960,
+  "max_window_layers": 36,
+  "model_type": "qwen3",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 36,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "rms_norm_eps": 1e-06,
+  "rope_parameters": {
+    "rope_theta": 1000000,
+    "rope_type": "default"
+  },
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "transformers_version": "5.13.0",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

generation_config.json

@@ -0,0 +1,13 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "5.13.0"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dec4e32ffbacfb7df6e69c89b7812600080141175b51b68342fb928bfe2851b
+size 16381517208

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
+size 11422650

tokenizer_config.json

@@ -0,0 +1,30 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": true,
+  "local_files_only": true,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}