Dataset Viewer
Auto-converted to Parquet Duplicate
id
stringlengths
9
11
md5
stringlengths
32
32
query
stringlengths
957
147k
images
listlengths
1
12
answer
dict
PMC10696902
d4ccc44a219a802606c9f5975d366237
You are analyzing a machine learning research paper. You will be given content from a paper, including: - method description - experimental setup - tables and figures related to the results Your task is to infer the main experimental conclusions that the authors draw from their results. Important: Do NOT simply desc...
[ "images/9225ee24984d.jpg", "images/45b2ea6822a0.jpg" ]
{ "reference_conclusions": [ "Therapeutic communities demonstrate significantly lower misconduct rates compared to other prison units with similar risk profiles, confirming the effectiveness of specialized unit environments in behavior management.", "Correctional officers in therapeutic communities utilize al...
PMC11694978
00cbc4109ffb06dadd8d7c936102d0eb
You are analyzing a machine learning research paper. You will be given content from a paper, including: - method description - experimental setup - tables and figures related to the results Your task is to infer the main experimental conclusions that the authors draw from their results. Important: Do NOT simply desc...
[ "images/1b74ef40604e.jpg" ]
{ "reference_conclusions": [ "The proposed MFGM method successfully identifies heterogeneous subgroups within functional data, revealing distinct dependency structures that homogeneous models would overlook.", "The EM algorithm combined with functional graphical lasso provides stable parameter estimation for ...
PMC11621163
046876af6649f06f311db8ccb8de0ec8
You are analyzing a machine learning research paper. You will be given content from a paper, including: - method description - experimental setup - tables and figures related to the results Your task is to infer the main experimental conclusions that the authors draw from their results. Important: Do NOT simply desc...
[ "images/e1b5234f8ff8.jpg", "images/29698935409a.jpg" ]
{ "reference_conclusions": [ "PFS management demonstrated exceptional safety with 100% survival in infants with gastroschisis, establishing its effectiveness as a primary treatment approach.", "Sepsis occurrence was a critical independent predictor of significantly prolonged hospital stays and extended parent...
PMC7616060
1e013a39a5604fdc774c99a858b1aff2
You are analyzing a machine learning research paper. You will be given content from a paper, including: - method description - experimental setup - tables and figures related to the results Your task is to infer the main experimental conclusions that the authors draw from their results. Important: Do NOT simply desc...
[ "images/5769301946eb.jpg" ]
{ "reference_conclusions": [ "Aromatic substrates specifically stabilize nitrilase helical filaments through interactions at oligomerization interfaces, while aliphatic nitriles and dinitriles do not promote filament formation.", "The active site is completely buried within each monomer, with substrate access...
PMC11707313
c534ca81e549525799cde740eb7a7818
You are analyzing a machine learning research paper. You will be given content from a paper, including: - method description - experimental setup - tables and figures related to the results Your task is to infer the main experimental conclusions that the authors draw from their results. Important: Do NOT simply desc...
[ "images/894d629d3f2d.jpg" ]
{ "reference_conclusions": [ "The e-cigarette starter kit intervention demonstrates clinically significant improvements in smoking cessation rates compared to standard signposting alone, validating its effectiveness as a smoking cessation strategy in emergency department settings.", "The intervention achieves...
PMC11599470
b63a447cb1a549e5fb450b5c324ac74d
You are analyzing a machine learning research paper. You will be given content from a paper, including: - method description - experimental setup - tables and figures related to the results Your task is to infer the main experimental conclusions that the authors draw from their results. Important: Do NOT simply desc...
[ "images/1fad976b7e97.jpg" ]
{ "reference_conclusions": [ "Higher basolateral amygdala activation to social reward cues amplifies the relationship between perceived social threat and subsequent social anxiety symptoms in adolescent girls.", "The interaction effect is specific to neural responses to social reward (not threat) cues, reveal...
PMC11705365
a548b162d5f0a803389ec7dc39f22f4d
You are analyzing a machine learning research paper. You will be given content from a paper, including: - method description - experimental setup - tables and figures related to the results Your task is to infer the main experimental conclusions that the authors draw from their results. Important: Do NOT simply desc...
[ "images/ae72ef6265e6.jpg", "images/076ce2c644ff.jpg", "images/eb258f9b06bd.jpg", "images/96227f802c0a.jpg", "images/3778a001d7dd.jpg" ]
{ "reference_conclusions": [ "Digital devices significantly strengthen public health surveillance capabilities during public health emergencies through multi-dimensional data collection, analysis, and dissemination.", "Public health surveillance serves as a critical mechanism for building city resilience, str...
PMC10697998
8b92bc50a6400e9e19feae10f399acdc
"You are analyzing a machine learning research paper.\n\nYou will be given content from a paper, inc(...TRUNCATED)
[ "images/d2616255780e.jpg" ]
{"reference_conclusions":["Keratoconus research has experienced exponential growth since 2008, with (...TRUNCATED)
PMC11709182
ca996d78af76bafd072416a618a7d989
"You are analyzing a machine learning research paper.\n\nYou will be given content from a paper, inc(...TRUNCATED)
[ "images/264fdc43ee3f.jpg", "images/9b5dedc1dd86.jpg", "images/69307c5d2a36.jpg" ]
{"reference_conclusions":["Recent hospitalization (3-6 months prior to home healthcare initiation) i(...TRUNCATED)
PMC11700113
79115af4c68f36c7f46723bb31753eee
"You are analyzing a machine learning research paper.\n\nYou will be given content from a paper, inc(...TRUNCATED)
[ "images/0ab7e840988b.jpg" ]
{"reference_conclusions":["CHB treatment programs in Eritrea face significant retention challenges w(...TRUNCATED)
End of preview. Expand in Data Studio

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Paper Conclusion RL Training

基于 EasyR1(verl)框架的论文结论强化学习训练,训练模型为 Qwen3-VL-8B-Thinking,使用外部 judge 模型(Qwen3-4B-Instruct-2507)对预测结论与 235B 教师模型参考结论进行打分。

目录结构

paper_conclusion_rl/
├── EasyR1/                                  # 训练框架(verl)
│   ├── verl/                                # 核心代码
│   ├── examples/
│   │   ├── paper_conclusion_grpo.yaml       # GRPO 训练配置
│   │   ├── format_prompt/
│   │   │   └── paper_conclusion_json.jinja  # prompt 模板
│   │   ├── reward_function/
│   │   │   ├── paper_conclusion_list_judge.py          # HTTP 模式 reward
│   │   │   ├── paper_conclusion_file_queue_judge.py    # 双机文件队列 reward
│   │   │   ├── paper_conclusion_judge_common.py        # 共用 judge 逻辑
│   │   │   └── file_queue_judge_worker.py              # 文件队列 worker
│   │   ├── qwen3_vl_8b_paper_conclusion_grpo.sh       # 训练启动脚本
│   │   ├── start_qwen3_4b_judge_vllm.sh               # judge 服务启动(HTTP)
│   │   └── start_qwen3_4b_judge_file_queue.sh         # judge 服务启动(双机)
│   ├── setup.py
│   ├── requirements.txt
│   └── ...
├── images_part00.tar                        # 图片分片压缩包(每个 ~2GB)
├── images_part01.tar
├── ...
├── paper_conclusion_rl_train.jsonl          # 训练集(4599 条)
├── paper_conclusion_rl_test.jsonl           # 测试集(1152 条)
└── README.md

注意:图片以 tar 分片形式存储,下载后需要先解压才能开始训练。

新机器环境准备

1. 硬件要求

  • 单机模式:至少 8 张 GPU(4 张跑 judge 服务,4 张跑训练)
  • 双机模式:每台机器至少 4 张 GPU(一台专跑训练,一台专跑 judge)
  • GPU 显存建议 >= 80GB(A100/H100),训练和 judge 同时运行时显存需求较高

2. 软件环境

# Python >= 3.10
python3 --version

# CUDA >= 12.1
nvidia-smi

3. 下载并解压数据集

# 下载数据集
huggingface-cli download <repo_id> --local-dir ./paper_conclusion_rl
cd paper_conclusion_rl

# 解压图片(解压后生成 images/ 目录)
for f in images_part*.tar; do tar xf "$f"; done

# 验证图片数量
ls images/ | wc -l  # 应该约 19659 个文件

4. 安装 EasyR1 依赖

cd EasyR1
pip install -e .

如果 pip install -e . 报错,可手动安装核心依赖:

pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121
pip install vllm>=0.8
pip install transformers accelerate
pip install hydra-core omegaconf
pip install pyarrow

5. 准备模型文件

需要下载两个模型到本地(不包含在此数据集中):

用途 模型 建议 HuggingFace 路径
RL 训练 Qwen3-VL-8B-Thinking Qwen/Qwen3-VL-8B-Thinking
Judge 服务 Qwen3-4B-Instruct-2507 Qwen/Qwen3-4B-Instruct-2507
# 示例:下载模型
huggingface-cli download Qwen/Qwen3-VL-8B-Thinking --local-dir /path/to/Qwen3-VL-8B-Thinking
huggingface-cli download Qwen/Qwen3-4B-Instruct-2507 --local-dir /path/to/Qwen3-4B-Instruct-2507

配置环境变量

训练脚本和 judge 脚本通过环境变量控制路径和行为,无需手动修改脚本内容

必须设置

# 项目根目录(即本 README 所在目录)
export PROJECT_ROOT=/path/to/paper_conclusion_rl

# 训练模型路径(新机器上的实际路径)
export MODEL_PATH=/path/to/Qwen3-VL-8B-Thinking

# Judge 模型路径(新机器上的实际路径)
# 注意:这个变量在 judge 启动脚本中名为 MODEL_PATH,需要在启动 judge 时设置

可选设置

# Judge 服务的 API key(如果 judge 需要鉴权,通常设为 EMPTY 即可)
export OPENAI_API_KEY=EMPTY

# CUDA 可见设备(默认脚本中有内置分配,也可手动覆盖)
# 单机模式示例:4卡 judge + 4卡训练
# export CUDA_VISIBLE_DEVICES=0,1,2,3  # judge 启动时
# export CUDA_VISIBLE_DEVICES=4,5,6,7  # 训练启动时

启动训练

方式一:单机 8 卡(4 卡 judge + 4 卡训练)

终端 1 — 启动 judge 服务:

cd $PROJECT_ROOT/EasyR1

CUDA_VISIBLE_DEVICES=0,1,2,3 \
MODEL_PATH=/path/to/Qwen3-4B-Instruct-2507 \
SERVED_MODEL_NAME=qwen3-4b-judge \
TP_SIZE=4 \
PORT=8000 \
bash examples/start_qwen3_4b_judge_vllm.sh

等待 judge 服务完全启动(看到 Uvicorn running on 日志)后,再启动训练。

终端 2 — 启动 RL 训练:

cd $PROJECT_ROOT/EasyR1

CUDA_VISIBLE_DEVICES=4,5,6,7 \
MODEL_PATH=/path/to/Qwen3-VL-8B-Thinking \
JUDGE_BASE_URL=http://127.0.0.1:8000/v1 \
JUDGE_MODEL=qwen3-4b-judge \
bash examples/qwen3_vl_8b_paper_conclusion_grpo.sh

方式二:双机(训练机和 judge 机分离)

适用于单机 GPU 显存不足的情况。两台机器需要能访问同一个共享文件系统(如 NFS)。

机器 B(Judge 机)— 启动 judge 服务 + 文件队列 worker:

cd $PROJECT_ROOT/EasyR1

QUEUE_ROOT=$PROJECT_ROOT/EasyR1/shared_judge_queue \
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
MODEL_PATH=/path/to/Qwen3-4B-Instruct-2507 \
TP_SIZE=8 \
bash examples/start_qwen3_4b_judge_file_queue.sh

机器 A(训练机)— 启动 RL 训练:

cd $PROJECT_ROOT/EasyR1

TRANSPORT_MODE=file_queue \
QUEUE_ROOT=$PROJECT_ROOT/EasyR1/shared_judge_queue \
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
TRAIN_GPUS_PER_NODE=8 \
ROLLOUT_TP_SIZE=8 \
MODEL_PATH=/path/to/Qwen3-VL-8B-Thinking \
bash examples/qwen3_vl_8b_paper_conclusion_grpo.sh

启动顺序:必须先启动 judge 机,确认 vLLM 服务和队列 worker 都就绪后,再启动训练机。否则训练会在等待 judge 结果时阻塞。

关键参数说明

训练配置(paper_conclusion_grpo.yaml

参数 默认值 说明
data.max_prompt_length 8192 最大 prompt 长度,超长样本会被过滤
data.max_response_length 2048 最大生成长度
data.rollout_batch_size 24 每 rollout 批次大小
worker.rollout.n 16 每个样本的 rollout 数量
worker.rollout.limit_images 12 每个样本最大图片数
worker.rollout.max_model_len 40960 rollout 模型最大上下文
worker.actor.model.lora.rank 32 LoRA 秩
worker.actor.optim.lr 1e-6 学习率
trainer.total_epochs 3 训练总轮数
trainer.save_freq 40 每 40 步保存 checkpoint
trainer.val_freq 40 每 40 步验证一次

Reward 计算

  • 使用 max(0, matched - wrong) / reference_count 公式
  • format_weight=0.05:格式正确性奖励权重
  • Judge 模型对每个样本的预测结论列表与 235B 教师参考结论进行逐一匹配打分

双机文件队列行为

使用双机模式时,训练和 judge 通过共享目录通信:

shared_judge_queue/
├── requests/
│   ├── pending/     # 训练写入请求,等待 judge 处理
│   └── processing/  # judge 正在处理的请求
└── results/
    ├── ok/          # 处理成功的结果
    └── error/       # 处理出错的结果
  • 确保 QUEUE_ROOT 在两台机器上指向同一个共享目录
  • 确保共享文件系统支持原子 rename 操作
  • 如果 requests/pending 中文件持续堆积,说明 judge worker 未正常消费
  • 如果 requests/processing 中文件持续堆积,说明 judge 服务可能已崩溃

数据集格式

每条训练样本格式:

{
    "id": "PMC12345",
    "md5": "abc123...",
    "query": "<prompt>\n\n<paper content with <image> placeholders>",
    "images": ["images/xxx.jpg", "images/yyy.jpg"],
    "answer": {
        "reference_conclusions": ["conclusion 1", "conclusion 2", ...],
        "rubrics": "...",
        "paper_id": "PMC12345",
        "md5": "abc123..."
    }
}
  • images 字段为相对路径,相对于项目根目录
  • reference_conclusions 来自 235B 教师模型,用于 judge 打分
  • 训练时学生模型只看到 queryimages,不直接看到参考结论

常见问题

Q: vLLM 报 AssertionError: Failed to apply prompt replacement for mm_items['image'][13]

A: 图片数超过 vLLM 处理上限。当前数据集已过滤掉图片数 > 12 的样本。如果仍有问题,降低 data.max_prompt_lengthworker.rollout.limit_images

Q: OOM(显存不足)

A: 尝试以下方法:

  • 降低 worker.rollout.gpu_memory_utilization(默认 0.5)
  • 减小 data.rollout_batch_size
  • 切换到双机模式,将 judge 和训练分到不同机器

Q: 训练阻塞在等待 reward

A: 检查 judge 服务是否正常运行。单机模式确认 http://127.0.0.1:8000/v1 可访问;双机模式检查 shared_judge_queue/ 中文件流转情况。

Q: 图片解压后找不到

A: 确保在项目根目录下执行 for f in images_part*.tar; do tar xf "$f"; done,解压后会生成 images/ 目录,与 JSONL 中的 images/xxx.jpg 路径对应。

Downloads last month
91