REVIEW 7 cited by
Analyzing and Reducing Catastrophic Forgetting in Parameter Efficient Tuning
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Analyzing and Reducing Catastrophic Forgetting in Parameter Efficient Tuning
read the original abstract
Existing research has shown that large language models (LLMs) exhibit remarkable performance in language understanding and generation. However, when LLMs are continuously fine-tuned on complex and diverse domain-specific downstream tasks, the inference performance on historical tasks decreases dramatically, which is known as a catastrophic forgetting problem. A trade-off needs to be kept between learning plasticity and memory stability. Plenty of existing works have explored strategies like memory replay, regularization and parameter isolation, but little is known about the geometric connection of various adjacent minima in the continual LLMs fine-tuning scenarios. In this work, we investigate the geometric connections of different minima through the lens of mode connectivity, which means different minima can be connected by a low-loss valley. Through extensive experiments, we uncover the mode connectivity phenomenon in the LLMs continual learning scenario and find that it can strike a balance between plasticity and stability. Building upon these findings, we propose a simple yet effective method called Interpolation-based LoRA (I-LoRA), which constructs a dual-memory experience replay framework based on LoRA parameter interpolations. Extensive experiments and analysis on eight domain-specific CL benchmarks demonstrate that I-LoRA consistently show significant improvement over the previous state-of-the-art approaches with up to $11\%$ performance gains, providing a strong baseline and insights for future research on the large language model continual learning problem. Our code is available at \url{https://github.com/which47/LLMCL}.
Forward citations
Cited by 7 Pith papers
-
FreeMOCA: Memory-Free Continual Learning for Malicious Code Analysis
FreeMOCA enables memory-free continual learning for malicious code analysis by adaptive layer-wise parameter interpolation between task updates, outperforming baselines on EMBER and AZ malware benchmarks with up to 42...
-
FreeMOCA: Memory-Free Continual Learning for Malicious Code Analysis
FreeMOCA enables memory-free continual learning for malicious code analysis via adaptive layer-wise interpolation between warm-started task optima, outperforming baselines on EMBER and AZ benchmarks with up to 42% acc...
-
Do Self-Evolving Agents Forget? Capability Degradation and Preservation in Lifelong LLM Agent Adaptation
Self-evolving LLM agents exhibit capability erosion under continual adaptation, which Capability-Preserving Evolution mitigates by raising retained simple-task performance from 41.8% to 52.8% in workflow evolution und...
-
Diffusion-Inspired Masked Fine-Tuning for Knowledge Injection in Autoregressive LLMs
Masked fine-tuning enables autoregressive LLMs to inject new factual knowledge without paraphrases and with reversal-curse resistance, matching diffusion LLM advantages on QA tasks.
-
Reward-Free Code Alignment from Pretrained or Fine-Tuned LLM: Unpacking the Trade-offs for Code Generation
Empirical study on five LLMs finds pretrained-to-aligned paths yield bigger gains over baseline than finetuned-to-aligned paths, though absolute accuracy remains lower for pretrained starts.
-
Securing LLMs in the Wild: Privacy and Security Challenges at the Edge
Edge LLM security is framed as a Security-Efficiency Paradox, with a three-wall constraint model, a composite SOES score, and a small FP/INT4 benchmark of six models.
-
Efficient Task Adaptation in Large Language Models via Selective Parameter Optimization
The paper claims a selective fine-tuning method that identifies and freezes core parameters to mitigate catastrophic forgetting in LLMs while improving domain adaptation, shown in experiments with GPT-J and LLaMA-3.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.