PulseAugur
EN
LIVE 10:37:05
tool · [1 source] ·

SymNoise fine-tuning method boosts LLaMA-2 performance by 6.7%

Researchers have introduced SymNoise, a novel method for instruction fine-tuning of language models that utilizes symmetric noise in embeddings. This technique aims to improve model performance by more strictly regulating local curvature. In evaluations using the LLaMA-2-7B model and the Alpaca dataset, SymNoise achieved a 69.04% score on AlpacaEval, a significant improvement over the state-of-the-art NEFTune method which scored 64.69%. The approach has also demonstrated consistent outperformance against NEFTune across various models and stronger instruction datasets. AI

Summary written by gemini-2.5-flash-lite from 1 sources. How we write summaries →

IMPACT Introduces a new fine-tuning technique that significantly improves language model performance on instruction-following tasks.

RANK_REASON The cluster contains an academic paper detailing a new method for fine-tuning language models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

COVERAGE [1]

  1. arXiv stat.ML TIER_1 · Abhay Yadav ·

    Understanding and Improving Noisy Embedding Techniques in Instruction Finetuning

    arXiv:2605.23171v1 Announce Type: cross Abstract: Recent advancements in instructional fine-tuning have injected noise into embeddings, with NEFTune (Jain et al., 2024) setting benchmarks using uniform noise. Despite NEFTune's empirical findings that uniform noise outperforms Gau…