Issues

• แนะนำ 11 สล็อต pg เว็บตรง แตกหนัก 2026 อันดับ 1 เชื่อถือได้ 999%

  simplescaling/s1#135 · Filter by author tedkuzmin22-beep opened on Apr 29, 2026

• Inconsistent AIME scores when porting evaluation to newer vLLM versions

  simplescaling/s1#134 · Filter by author asparius opened on Mar 9, 2026
  3

• Cannot find simplescaling/s1-7B available / renamed / removed?

  simplescaling/s1#133 · Filter by author SondosBsharat opened on Feb 18, 2026
  1

• Performance on Qwen2.5-7B-Instruct

  simplescaling/s1#132 · Filter by author kyre-99 opened on Feb 14, 2026
  1

• DeepSeek-R1 trace generation details

  simplescaling/s1#130 · Filter by author Narabzad opened on Jan 14, 2026
  1

• I tried to finetune qwen3-1.7B on implescaling/s1K-1.1_tokenized datasets , by using this repository code , cannot figure out why do I got worse inferrence result ?

  simplescaling/s1#128 · Filter by author hzgdeer opened on Oct 21, 2025
  6

• does the budget forcing works?

  simplescaling/s1#127 · Filter by author maziye-azad opened on Oct 2, 2025
  4

• final answer extraction from the solution for aime24 dataset

  simplescaling/s1#126 · Filter by author maziye-azad opened on Oct 2, 2025
  1

• Thanks for your work, can I use "Qwen/Qwen2.5-32B-Instruct" instead of gpt-4o-mini for evaluation? If yes, what codes need to be changed?

  simplescaling/s1#124 · Filter by author LiZhangMing opened on Jul 21, 2025
  3

• Performance decrease after finetuning on 8*A800 GPUs. Got 0.367 in aime24_nofigures task

  simplescaling/s1#123 · Filter by author LiZhangMing opened on Jun 26, 2025
  16

• Performance Drop on MGSM After Finetuning - QWEN2.5 - 3B

  simplescaling/s1#122 · Filter by author baselmousi opened on Jun 24, 2025
  7

• Evaluation of local fine-tuned checkpoint is ~10× slower than the same-weight original Hugging Face model

  simplescaling/s1#118 · Filter by author SondosBsharat opened on May 10, 2025
  1

• budget forcing

  simplescaling/s1#117 · Filter by author Kenwwww opened on May 9, 2025

• 1.5B accuracy 0 after SFT

  simplescaling/s1#115 · Filter by author alvinshao0313 opened on Apr 19, 2025
  8

• what is the content of "results/gemini/{subdir}/*.json"

  simplescaling/s1#114 · Filter by author Pyjacc opened on Apr 18, 2025
  1

• Is the code data used in the training data?

  simplescaling/s1#112 · Filter by author xuhu0115 opened on Apr 8, 2025
  4

• Training time shows 18 hr on 16 * H100

  simplescaling/s1#111 · Filter by author kartikjain42 opened on Apr 4, 2025
  3

• Config for smaller models

  simplescaling/s1#110 · Filter by author DoubtedSteam opened on Apr 1, 2025
  1

• Why generate more than 8k tokens at a time?

  simplescaling/s1#109 · Filter by author JaydencoolCC opened on Apr 1, 2025
  1

• Issue Reproducing s1.1-32B Training Loss (Observed vs. WandB)

  simplescaling/s1#108 · Filter by author dzh19990407 opened on Mar 31, 2025
  10

• Can I specify the thinking context of the query when s1 inferencing with vllm?

  simplescaling/s1#107 · Filter by author Siki-cloud opened on Mar 24, 2025
  1

• qwen2.5-32B can not fine tuning on 80GB A100 in max_seq_length=32768

  simplescaling/s1#106 · Filter by author Gresham429 opened on Mar 21, 2025
  5

• why the model file size is 129g?

  simplescaling/s1#104 · Filter by author wikithink opened on Mar 20, 2025
  1

• The minimum GPU resources needed to fine-tune the 32B model?

  simplescaling/s1#103 · Filter by author JaydencoolCC opened on Mar 19, 2025
  1

• Error in Evaluating on Other Dataset (AssertionError: min_tokens_thinking only supports until_thinking tokens that are 1 token long)

  simplescaling/s1#102 · Filter by author jd730 opened on Mar 19, 2025
  3