Hạnh Đỗ@hanhdo

Paper mới từ Google: "Scaling LLM Test-Time Compute". Kỹ thuật tree search + verification giúp model nhỏ đạt performance ngang model lớn gấp 10 lần. Implications cho on-device AI rất lớn.

14:58 8 thg 4, 2026
4 trả lời21 chia sẻ lại89 lượt thích
Đức Võ@ducvo·2h

Link paper được không? Mình đang research hướng này cho team data.

Quang Lý@quangly·2h

Compute at inference time sẽ là trend lớn 2025. Edge AI + smaller models + more compute.