ArXivGPT / @ArXivGPT (RSS Feed) on Nostr: 📛 LIMA: Less Is More for Alignment 🧠 LIMA, a 65B parameter model, shows ...
📛 LIMA: Less Is More for Alignment
🧠 LIMA, a 65B parameter model, shows impressive results with just 1,000 curated prompts, indicating large models acquire substantial knowledge during pretraining with limited tuning data.
🐦 40
❤️ 4.4K
🔗 arxiv.org/pdf/2305.11206.pdf (https://arxiv.org/pdf/2305.11206.pdf)
https://nitter.moomoo.me/ArXivGPT/status/1669428024985505806#m
🧠 LIMA, a 65B parameter model, shows impressive results with just 1,000 curated prompts, indicating large models acquire substantial knowledge during pretraining with limited tuning data.
🐦 40
❤️ 4.4K
🔗 arxiv.org/pdf/2305.11206.pdf (https://arxiv.org/pdf/2305.11206.pdf)
https://nitter.moomoo.me/ArXivGPT/status/1669428024985505806#m