Unverified50% confidenceFactExact time
The study trained autoregressive sequence models on 1000 seed molecules, compared canonical SAFE with randomized SAFE across increasing corpus sizes, and isolated the two randomization axes of atom order and fragment order.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedByteDance's Seed team plans to train a foundation large model with more than 50 trillion parameters; the project is currently in the early validation stage and insists on native pre-training while rejecting the distillation approach.72% similarUnverifiedIn low-data molecule generation, prioritizing representations that support fragment-level randomization (such as SAFE) and enabling both atom-order and fragment-order randomization is more effective than choosing only one.71% similarUnverifiedIn stochastic machine learning experiments, the result of each run is affected by two sources of randomness: the random seed and the evaluation items.71% similarUnverifiedWhen conducting single-edit comparison experiments, fixing the seed maintains a consistent baseline and isolates variables.70% similarUnverifiedMasked diffusion randomly masks some tokens in the sequence during training, and during generation starts from the masked state and gradually denoises through multiple iterations to restore the complete sequence, with each step able to predict tokens at multiple positions in parallel.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114336API
curl https://kongchang.com/api/v1/knowledge/claims/1114336MCP
get_claim(id=1114336)