Context-Free Synthetic Data Mitigates Forgetting
Fine-tuning a language model often results in a degradation of its existing performance on other tasks, due to a shift in the model parameters; this phenomenon is often referred to as (catastrophic) forgetting. We are interested in mitigating this, in settings where we only have access to the model weights but no access to its training data/recipe. A natural approach is to penalize the KL divergence between the original model and the new one. Our main realization is that a simple process - which we term context-free generation - allows for an approximate unbiased estimation of this KL divergence. We show that augmenting a fine-tuning dataset with context-free generations mitigates forgetting, in two settings: (a) preserving the zero-shot performance of pretrained-only models, and (b) preserving the reasoning performance of thinking models. We show that contextual synthetic data, and even a portion of the pretraining data, are less effective. We also investigate the effect of choices like generation temperature, data ratios etc. We present our results for OLMo-1B for pretrained-only setting and R1-Distill-Llama-8B for the reasoning setting.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
Towards Lifelong Few-Shot Customization of Text-to-Image Diffusion
Lifelong few-shot customization for text-to-image diffusion aims to continually generalize existing models for new tasks with minimal data while preserving old knowledge. Current customization diffusion models excel in f…
Data-free Knowledge DistillationKnowledge DistillationSER-Diff: Synthetic Error Replay Diffusion for Incremental Brain Tumor Segmentation
Incremental brain tumor segmentation is critical for models that must adapt to evolving clinical datasets without retraining on all prior data. However, catastrophic forgetting, where models lose previously acquired know…
Brain Tumor SegmentationKnowledge DistillationIncremental LearningSinkTrack: Attention Sink based Context Anchoring for Large Language Models
Large language models (LLMs) suffer from hallucination and context forgetting. Prior studies suggest that attention drift is a primary cause of these problems, where LLMs' focus shifts towards newly generated tokens and …
A Multi-Encoder Frozen-Decoder Approach for Fine-Tuning Large Language Models
Among parameter-efficient fine-tuning methods, freezing has emerged as a popular strategy for speeding up training, reducing catastrophic forgetting, and improving downstream performance. We investigate the impact of fre…
Decoderparameter-efficient fine-tuningParameter Alignment Mitigates Catastrophic Forgetting in Multilingual Expert Language Models
While continual pretraining~(CPT) is a practical way to extend large language models to new languages, naïve finetuning on targeted data erodes existing capabilities through catastrophic forgetting. Organizing training a…
Reading ComprehensionContinual PretrainingLanguage AcquisitionGeneral Knowledge