Optimizer-Model Consistency: Full Finetuning with the Same Optimizer as Pretraining Forgets Less

📰 ArXiv cs.AI

Using the same optimizer for pretraining and finetuning can reduce forgetting in large language models, leading to better performance on new tasks

advanced Published 9 May 2026
Action Steps
  1. Choose a suitable optimizer for pretraining a large language model
  2. Use the same optimizer for full finetuning on a new task
  3. Compare the performance of the model with and without optimizer consistency
  4. Evaluate the tradeoff between learning and forgetting in the model
  5. Apply the optimizer consistency technique to achieve better results on downstream tasks
Who Needs to Know This

Researchers and engineers working with large language models can benefit from this insight to improve their model's performance and reduce forgetting, especially when finetuning on new tasks

Key Insight

💡 Optimizer-model consistency can improve the learning-forgetting tradeoff in large language models

Share This
🚀 Using the same optimizer for pretraining & finetuning can reduce forgetting in LLMs! 🤖

Full Article

Title: Optimizer-Model Consistency: Full Finetuning with the Same Optimizer as Pretraining Forgets Less

Abstract:
arXiv:2605.06654v1 Announce Type: cross Abstract: Optimizers play an important role in both pretraining and finetuning stages when training large language models (LLMs). In this paper, we present an observation that full finetuning with the same optimizer as in pretraining achieves a better learning-forgetting tradeoff, i.e., forgetting less while achieving the same or better performance on the new task, than other optimizers and, possibly surprisingly, LoRA, during the supervised finetuning (SF
Read full paper → ← Back to Reads

Related Videos

5 Levels of AI Agents - From Simple LLM Calls to Multi-Agent Systems
5 Levels of AI Agents - From Simple LLM Calls to Multi-Agent Systems
Dave Ebbelaar (LLM Eng)
10 NotebookLM Features You Missed & You Shouldn't
10 NotebookLM Features You Missed & You Shouldn't
Educraft
ChatGPT Images 2.0 is Insane For Infographics (Sorry Gemini)
ChatGPT Images 2.0 is Insane For Infographics (Sorry Gemini)
Educraft
The best LLM is..... (2nd Qtr 2025 latest ranking)
The best LLM is..... (2nd Qtr 2025 latest ranking)
AI Mastermind
How to Use ChatGPT’s Deep Research to Save HOURS on Research
How to Use ChatGPT’s Deep Research to Save HOURS on Research
AI Mastermind
The Best LLM Is.... (A breakdown for every category)
The Best LLM Is.... (A breakdown for every category)
AI Mastermind