Unverified50% confidenceFactExact time
RecSA is initialized from Qwen3-4B and trained on 0.62B tokens at the original model's 32K training length.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/6/2026
First Seen
Valid until: 2/4/2027
Sources
Related Entities
Related Claims
UnverifiedUsing QLoRA on A100 and RTX PRO 6000, training speed for Qwen3.5-35B-A3B can be improved by up to 4.1×, and for Qwen3-30B-A3B by up to 3.3×.68% similarUnverified340M and 1.3B are parameter scales, and 4K context means a training/evaluation sequence length of 4096 tokens.65% similarUnverifiedAccordion-6B-A0.5B was pre-trained on 3.6T tokens.65% similarUnverifiedAccording to the Chinchilla scaling law, the optimal training data volume for a 30B model is approximately 600B–1T tokens.65% similarUnverified3,189 successful rollouts were collected from Qwen3.8-27B for supervised fine-tuning (imitation learning), and performance stagnated at 47.5%, below the results of two rounds of RL training.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1141963API
curl https://kongchang.com/api/v1/knowledge/claims/1141963MCP
get_claim(id=1141963)