README.md

---
dataset_info:
- config_name: Qwen3-VL-4B-Thinking-gen-4096
  features:
  - name: source
    dtype: string
  - name: item_id
    dtype: string
  - name: origin_id
    dtype: int64
  - name: sample_seed
    dtype: int64
  - name: question
    dtype: string
  - name: images
    list: image
  - name: num_images
    dtype: int32
  - name: ref_answer
    dtype: string
  - name: category
    dtype: string
  - name: input_ids
    list: int64
  - name: input_tokens_len
    dtype: int64
  - name: generated_texts
    dtype: string
  - name: generated_tokens_len
    dtype: int64
  - name: n_prefill_tokens
    dtype: int64
  - name: finish_reason
    dtype: string
  - name: seed
    dtype: int64
  - name: labels
    list: int64
  - name: image_fingerprint
    dtype: string
  - name: image_processor_config
    dtype: string
  - name: image_processor_config_hash
    dtype: string
  - name: sum_vision_tokens
    dtype: int64
  - name: max_vision_tokens
    dtype: int64
  - name: id
    dtype: string
  - name: truncated_by_cap
    dtype: bool
  - name: generation_model_name_or_path
    dtype: string
  - name: enable_thinking
    dtype: bool
  splits:
  - name: train
    num_bytes: 34791348
    num_examples: 303
  download_size: 32146015
  dataset_size: 34791348
- config_name: Qwen3-VL-4B-Thinking-gen-4096-lmms-eval-q
  features:
  - name: source
    dtype: string
  - name: item_id
    dtype: string
  - name: origin_id
    dtype: int64
  - name: sample_seed
    dtype: int64
  - name: question
    dtype: string
  - name: images
    list: image
  - name: num_images
    dtype: int32
  - name: ref_answer
    dtype: string
  - name: category
    dtype: string
  - name: input_ids
    list: int64
  - name: input_tokens_len
    dtype: int64
  - name: generated_texts
    dtype: string
  - name: generated_tokens_len
    dtype: int64
  - name: n_prefill_tokens
    dtype: int64
  - name: finish_reason
    dtype: string
  - name: seed
    dtype: int64
  - name: labels
    list: int64
  - name: image_fingerprint
    dtype: string
  - name: image_processor_config
    dtype: string
  - name: image_processor_config_hash
    dtype: string
  - name: sum_vision_tokens
    dtype: int64
  - name: max_vision_tokens
    dtype: int64
  - name: id
    dtype: string
  - name: truncated_by_cap
    dtype: bool
  - name: generation_model_name_or_path
    dtype: string
  - name: enable_thinking
    dtype: bool
  splits:
  - name: train
    num_bytes: 35588114
    num_examples: 304
  download_size: 32790185
  dataset_size: 35588114
configs:
- config_name: Qwen3-VL-4B-Thinking-gen-4096
  data_files:
  - split: train
    path: Qwen3-VL-4B-Thinking-gen-4096/train-*
- config_name: Qwen3-VL-4B-Thinking-gen-4096-lmms-eval-q
  data_files:
  - split: train
    path: Qwen3-VL-4B-Thinking-gen-4096-lmms-eval-q/train-*
---