Upload folder using huggingface_hub

.gitattributes

@@ -1,35 +1,36 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz 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
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz 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
+assets/fig1_aime24_curves_added.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,9 +1,188 @@
 ---
-license: apache-2.0
+license: mit
+library_name: transformers
 datasets:
 - BytedTsinghua-SIA/DAPO-Math-17k
 language:
 - en
 base_model:
 - deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
----
+pipeline_tag: text-generation
+---
+
+<div align="center">
+<span style="font-family: default; font-size: 1.5em;">AscentRL: Simplicity at Scale</span>
+<div>
+🚀 Competitive RL Performance Without Complex Techniques 🌟
+</div>
+</div>
+
+<br>
+
+<div align="center" style="line-height: 1;">
+  <a href="[YOUR_GITHUB_REPO]" style="margin: 2px;">
+    <img alt="Code" src="https://img.shields.io/badge/GitHub-100000?style=for-the-badge&logo=github&logoColor=white" style="display: inline-block; vertical-align: middle;"/>
+  </a>
+  <a href="[YOUR_BLOG_LINK]" target="_blank" style="margin: 2px;">
+    <img alt="Notion" src="https://img.shields.io/badge/Notion-%23000000.svg?style=for-the-badge&logo=notion&logoColor=white" style="display: inline-block; vertical-align: middle;"/>
+  </a>
+</div>
+
+</div>
+</div>
+
+## Overview
+
+**AscentRL** demonstrates that competitive reinforcement learning performance for small language models doesn't require complex multi-stage pipelines or sophisticated stabilization techniques. Using a minimal recipe with single-stage training and fixed hyperparameters, we achieve state-of-the-art results on mathematical reasoning tasks.
+
+We release two models:
+- **AscentRL-1.5B-Weak**: Trained from DeepSeek-R1-Distill-Qwen-1.5B
+- **AscentRL-1.5B-Strong**: Trained from OpenMath-Nemotron-1.5B
+
+Both models use identical hyperparameters without per-model tuning, demonstrating the robustness of our approach.
+
+![The AIME24 performance curve for scaling from a weak base DeekSeek-R1-Distill-Qwen-1.5B and a strong base OpenMath-Nemotron-1.5B over thousands of steps.](./assets/fig1_aime24_curves_added.png)
+
+## Key Highlights
+
+✨ **Simplicity**: Single-stage training with fixed hyperparameters—no multi-stage pipelines, no dynamic schedules, no specialized stabilization techniques
+
+📈 **Stability**: Smooth, monotonic improvement over 4,000+ training steps without collapses or oscillations
+
+🎯 **Performance**: State-of-the-art results at 1.5B scale, matching or exceeding more complex approaches
+
+💰 **Efficiency**: Comparable or better performance with less compute than multi-stage methods
+
+🔓 **Open**: Complete evaluation scripts, and model weights released
+
+## Performance
+
+### AscentRL-1.5B-Weak (Based on DeepSeek-R1-Distill-Qwen-1.5B)
+
+| Model                    | AIME24 (@32) | AIME25 (@32) | AMC23 (@32) | MATH-500 (@4) | Minerva (@4) | OlympiadBench (@4) | HMMT25 (@32) | BRUMO25 (@32) | CMIMC25 (@32) | Avg       |
+| ------------------------ | ------------ | ------------ | ----------- | ------------- | ------------ | ------------------ | ------------ | ------------- | ------------- | --------- |
+| DeepSeek-R1-Distill-1.5B | 29.90        | 22.40        | 63.82       | 84.90         | 34.65        | 45.95              | 13.44        | 30.94         | 12.89         | 37.65     |
+| DeepScaleR-1.5B-Preview  | 40.21        | 28.65        | 73.83       | 89.30         | 39.34        | 52.79              | 18.96        | 40.00         | 21.00         | 44.88     |
+| ProRL-V2                 | 51.87        | 35.73        | 88.75       | 92.00         | 49.03        | **67.84**          | 19.38        | 47.29         | **25.86**     | 53.08     |
+| BroRL                    | **57.50**    | 36.88        | /           | **92.14**     | 49.08        | 61.54              | /            | /             | /             | /         |
+| AscentRL-1.5B-Weak       | 52.29        | **37.19**    | **91.02**   | 91.55         | **51.47**    | 66.77              | **21.98**    | **52.71**     | 25.63         | **54.51** |
+
+Besides, the real question is whether our simplicity comes at a computational cost. It doesn't. We match half of ProRL-V2's compute budget while using a single-stage recipe with fixed hyperparameters. BroRL requires 4.9× more compute by increasing rollouts to 512 per example, essentially exhaustively exploring the solution space. Our approach achieves competitive performance without this computational overhead.
+
+|                              | w/ Dynamic Sampling?            | Training Steps | Train Batch Size | Rollout N    | Max Context Length | Estimated Total Token Budget                                 |
+| ---------------------------- | ------------------------------- | -------------- | ---------------- | ------------ | ------------------ | ------------------------------------------------------------ |
+| DeepScaleR-1.5B-Preview      | ❌                               | 1,750          | 128              | 8            | 8k → 16k → 24k     | $(1040\times8k + 480\times16k + 230\times24k) \times 128\times 8 \approx2.2\times10^6k$ |
+| ProRL-V1                     | ✅   Filter Ratio $\approx 50\%$ | 2,450          | 256              | 16 → 32 → 16 | 8k → 16k           | $\frac 1 {50\%}(1700\times16\times8k + 550\times32\times8k + 200\times16\times16k) \times 256\approx2.1\times10^8k$ |
+| ProRL-V2 (Based on ProRL-V1) | ✅   Filter Ratio $\approx 50\%$ | +1,000         | 256              | 16 → 32 → 16 | 8k → 16k → 8k      | $\frac 1 {50\%}(1700\times16\times8k + 550\times32\times8k + 200\times16\times16k + 1000\times16\times8k) \times 256\approx2.8\times10^8k$ |
+| BroRL (Based on ProRL-V2)    | ✅   Filter Ratio $\approx 50\%$ | +191           | 128              | 512          | 16k                | $\frac 1 {50\%}[(1700\times16\times8k + 550\times32\times8k + 200\times16\times16k + 1000\times16\times8k) \times 256 + 191\times512\times16k\times128]\approx6.8\times10^8k$ |
+| AscentRL-1.5B-Weak           | ❌                               | 4,380          | 256              | 8            | 16k                | $4380\times256\times 8\times 16k \approx1.4\times10^8k$      |
+
+**Note on dynamic sampling**: Models marked with ✅ use dynamic sampling to filter examples. Following [POLARIS](https://honorable-payment-890.notion.site/POLARIS-A-POst-training-recipe-for-scaling-reinforcement-Learning-on-Advanced-ReasonIng-modelS-1dfa954ff7c38094923ec7772bf447a1), we estimate a 50% filter ratio for DeepSeek-R1-Distill-Qwen-1.5B using dynamic sampling, as rollouts often contain many trivial/hard cases (e.g., 8/8 or 0/8 correct rollouts). Even assuming no filtering (i.e., 0% ratio), our compute use remains comparable or even lower, making our estimates conservative.
+
+### AscentRL-1.5B-Strong (Based on OpenMath-Nemotron-1.5B)
+
+| Model                  | AIME24 (@32) | AIME25 (@32) | AMC23 (@32) | MATH-500 (@4) | Minerva (@4) | OlympiadBench (@4) | HMMT25 (@32) | BRUMO25 (@32) | CMIMC25 (@32) | Avg       |
+| ---------------------- | ------------ | ------------ | ----------- | ------------- | ------------ | ------------------ | ------------ | ------------- | ------------- | --------- |
+| OpenMath-Nemotron-1.5B | 58.75        | 48.44        | 90.55       | 92.40         | 26.93        | 71.70              | 30.10        | 61.67         | 30.08         | 56.74     |
+| QUESTA-Nemotron-1.5B   | **71.56**    | 62.08        | 93.44       | 92.95         | **32.08**    | 72.28              | **40.94**    | **67.50**     | 41.48         | 63.81     |
+| AscentRL-1.5B-Strong   | 69.69        | **62.92**    | **96.02**   | **94.15**     | 30.24        | **76.59**          | 40.63        | 66.88         | **41.72**     | **64.32** |
+
+We achieve 64.32% average, slightly outperforming QuestA's 63.81% and leading on five of nine benchmarks. The gap is narrow, which makes sense—both approaches are pushing the boundaries of what's achievable at 1.5B scale. The key difference is in how we get there. We use less compute while achieving slightly better average performance without designing a complex curriculum as used in QuestA.
+
+|                      | w/ Dynamic Sampling?             | Training Steps | Train Batch Size | Rollout N | Max Context Length | Estimated Total Token Budget                                 |
+| -------------------- | -------------------------------- | -------------- | ---------------- | --------- | ------------------ | ------------------------------------------------------------ |
+| QUESTA-Nemotron-1.5B | ✅    Filter Ratio $\approx 50\%$ | 2,000          | 128              | 16        | 32k                | $\frac 1 {50\%}\times2000\times128 \times 16\times 32k \approx2.6\times10^8k$ |
+| AscentRL-1.5B-Strong | ❌                                | 3,440          | 256              | 8         | 16k                | $3440\times256 \times 8\times 16k \approx1.1\times10^8k$     |
+
+## Training Recipe
+
+Our approach is deliberately minimal:
+
+**Core Algorithm**: Standard GRPO with binary outcome rewards
+- **Reward**: Simple DAPO verifier (string-matching, no SymPy)
+- **Training**: Single-stage, no curriculum or stage transitions
+- **Hyperparameters**: Fixed throughout (no adaptive schedules)
+- **Data**: DAPO-Math-17k without filtering or dynamic sampling
+- **Length Control**: 16K context cap (no explicit penalties)
+- **Stabilization**: Only "clip higher" for gradient stability
+
+Detail hyperparameters and comparisons on training techniques with other methods can refer to our blog.
+
+## Training Data
+
+We train on [DAPO-Math-17k](https://huggingface.co/datasets/BytedTsinghua-SIA/DAPO-Math-17k), a curated dataset of mathematical problems. **No offline difficulty filtering or online dynamic sampling is used.**
+
+## Usage
+
+### Basic Inference
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model_name = "hbx/AscentRL-1.5B-Strong"  # or AscentRL-1.5B-Weak
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    torch_dtype="auto",
+    device_map="auto"
+)
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+prompt = """<problem>
+
+Please reason step by step, and put your final answer within \\boxed{}."""
+
+messages = [{"role": "user", "content": prompt}]
+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,
+    temperature=0.7,
+    top_p=0.9,
+    do_sample=True
+)
+
+response = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
+print(response)
+```
+
+### Batch Inference with vLLM
+```python
+from vllm import LLM, SamplingParams
+
+llm = LLM(
+    model="hbx/AscentRL-1.5B-Strong",
+    tensor_parallel_size=1,
+    max_model_len=32768
+)
+
+sampling_params = SamplingParams(
+    temperature=0.7,
+    top_p=0.9,
+    max_tokens=16384,
+)
+
+problems = [...]  # Your list of problems
+responses = llm.generate(problems, sampling_params)
+```
+
+## Reproduction
+
+We provide evaluation scripts based on [POLARIS](https://github.com/ChenxinAn-fdu/POLARIS), the evaluation script is [TODO](TODO).
+
+## Citation
+
+```bibtex
+@misc{he2025ascentrl,
+  title        = {TODO},
+  author       = {TODO},
+  year         = {2025},
+  month        = {Nov},
+  day          = {1},
+  note         = {First published on Notion},
+  url          = {https://TODO}
+}

config.json

@@ -0,0 +1,29 @@
+{
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "hidden_act": "silu",
+  "hidden_size": 1536,
+  "initializer_range": 0.02,
+  "intermediate_size": 8960,
+  "max_position_embeddings": 131072,
+  "max_window_layers": 21,
+  "model_type": "qwen2",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 2,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 10000,
+  "sliding_window": 4096,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": true,
+  "use_mrope": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

generation_config.json

@@ -0,0 +1,9 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151646,
+  "eos_token_id": 151643,
+  "do_sample": true,
+  "temperature": 0.6,
+  "top_p": 0.95,
+  "transformers_version": "4.39.3"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1bedcdf243c4e2e633fa04c05617809cf6a1bbc1a07221609035f6347efecffb
+size 3554214752

tokenizer_config.json

@@ -0,0 +1,35 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "bos_token": {
+    "__type": "AddedToken",
+    "content": "<｜begin▁of▁sentence｜>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "clean_up_tokenization_spaces": false,
+  "eos_token": {
+    "__type": "AddedToken",
+    "content": "<｜end▁of▁sentence｜>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "legacy": true,
+  "model_max_length": 16384,
+  "pad_token": {
+    "__type": "AddedToken",
+    "content": "<｜end▁of▁sentence｜>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sp_model_kwargs": {},
+  "unk_token": null,
+  "tokenizer_class": "LlamaTokenizerFast",
+  "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% set ns = namespace(is_first=false, is_tool=false, is_output_first=true, system_prompt='') %}{%- for message in messages %}{%- if message['role'] == 'system' %}{% set ns.system_prompt = message['content'] %}{%- endif %}{%- endfor %}{{bos_token}}{{ns.system_prompt}}{%- for message in messages %}{%- if message['role'] == 'user' %}{%- set ns.is_tool = false -%}{{'<｜User｜>' + message['content']}}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is none %}{%- set ns.is_tool = false -%}{%- for tool in message['tool_calls']%}{%- if not ns.is_first %}{{'<｜Assistant｜><｜tool▁calls▁begin｜><｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{%- set ns.is_first = true -%}{%- else %}{{'\\n' + '<｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{{'<｜tool▁calls▁end｜><｜end▁of▁sentence｜>'}}{%- endif %}{%- endfor %}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is not none %}{%- if ns.is_tool %}{{'<｜tool▁outputs▁end｜>' + message['content'] + '<｜end▁of▁sentence｜>'}}{%- set ns.is_tool = false -%}{%- else %}{% set content = message['content'] %}{% if '</think>' in content %}{% set content = content.split('</think>')[-1] %}{% endif %}{{'<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>'}}{%- endif %}{%- endif %}{%- if message['role'] == 'tool' %}{%- set ns.is_tool = true -%}{%- if ns.is_output_first %}{{'<｜tool▁outputs▁begin｜><｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- set ns.is_output_first = false %}{%- else %}{{'\\n<｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- endif %}{%- endif %}{%- endfor -%}{% if ns.is_tool %}{{'<｜tool▁outputs▁end｜>'}}{% endif %}{% if add_generation_prompt and not ns.is_tool %}{{'<｜Assistant｜><think>\\n'}}{% endif %}"
+}