#Llama-2-7B

2 stories · 0 in 7 days · 1 in 30 days

BeaverTails 1 DARTS 1 VaccineBooster 1 对齐 1 微调安全 1 模型合并 1 表示调优 1 解码器 1

Stories, products and papers tagged "Llama-2-7B", newest first.

9月30日

Papers BuildersMulti-source09:17

Chinese contentVaccineBooster：混合防御提升模型对有害微调的鲁棒性

OpenAI新研究结合两种防御技术，让模型在保持拒绝能力的同时减少有害内容输出。

#VaccineBooster #Llama-2-7B #BeaverTails #对齐

Event

arXiv cs.LG@Muhammad Zeeshan Akram et al. (5)11 sourcesSource

8月31日

PapersSelected73°11:41

DARTS: Decoder-Aware Representation Tuning via Surgery

DARTS improves multi-task model merging for decoder LLMs with minimal parameter overhead, benefiting developers working with code generation and reasoning models.

#DARTS #Llama-2-7B #model merging #representation bias

arXiv cs.LG@Aaryan Ajay Sharma et al. (3)Source