Skip to content
Quantoros

Your fine-tuned LLM forgot its general capabilities (catastrophic forgetting). How do you fix it?

LLM & AILevel: IntermediateTroubleshooting Scenarios

Your fine-tuned LLM forgot its general capabilities (catastrophic forgetting). How do you fix it?

Hints
  • Which parameters change and which stay frozen?
  • Separate memory for weights, gradients, optimizer states and activations.

Answer

  1. Switch to LoRA — if you used full fine-tuning, the primary fix is using LoRA; frozen base weights cannot forget
  2. Add data mixing — re-run fine-tuning with 10–20% general-domain data mixed in
  3. Reduce training intensity — fewer epochs, lower learning rate, reduce LoRA rank
  4. Merge carefully — if merging LoRA into base, use a lower scaling factor (α/r ratio)
  5. Evaluate incrementally — track both task metric and general benchmarks during training; apply early stopping when general capability starts degrading

Source and licence: JustInCache & contributors — Awesome AI Interviews · Apache-2.0

Related questions