dvlab-research/Step-DPO

Issues

• Ablation between DPO and Step-DPO

  #20 opened 2 months ago by tqzhong
  1

• Question about eval results on Qwen2-7B-Instruct

  #21 opened 19 days ago by LUOMU17
  0

• Request for Citation

  #5 opened 5 months ago by hbin0701
  3

• Does step-dpo work?

  #19 opened 2 months ago by hxdtest
  0

• question about StepDPOTrainer

  #18 opened 2 months ago by FlyingDutchman26
  1

• eval_math:143 prompt_answer = remove_text(prompt_answer)

  #17 opened 3 months ago by Xalp
  0

• question about Data Construction

  #15 opened 3 months ago by hong-xl
  1

• I followed the steps in the README file to train the model, but I got an error. Here is the error message.

  #16 opened 3 months ago by Claude121381011
  0

• 复现问题

  #2 opened 5 months ago by yyht
  8

• Evaluation scripts for AIME and Odyssey-MATH

  #14 opened 3 months ago by bmanczak
  0

• share sft-dataset

  #1 opened 5 months ago by yyht
  4

• deepseek-math-7b-rl-stepdpo推理后的结果问题

  #13 opened 4 months ago by wjn1996
  1

• Great work， what about the computation resources needed for each experiment

  #6 opened 5 months ago by yanghu819
  4

• validation set

  #12 opened 4 months ago by kaishxu
  0

• appendix missing

  #11 opened 4 months ago by ChrisMii
  1

• During DPO training, will SFT loss be calculated?

  #10 opened 4 months ago by mohhao
  0

• question about Data Construction Pipeline

  #9 opened 4 months ago by yyht
  0

• questions about some parameter in config_full.yaml

  #8 opened 4 months ago by kaishxu
  1

• Data Generation Pipeline

  #4 opened 5 months ago by yapdianang
  2

• Question about the DPO vs. Step-DPO.

  #7 opened 5 months ago by flow3rdown
  2

• About details of Step localization and Rectification

  #3 opened 5 months ago by ToheartZhang
  1