Meta, MIT and UW researchers make a model that rewrites its context
The model rewrites its context, keeps important facts, and removes data that is not important.
Claimed, not confirmed
Researchers from Meta, MIT, and the University of Washington made a new type of language model, the Context Language Model (CLM). A CLM rewrites its context, keeps important facts, and removes data that is not important. Other methods use rules that do not change to decrease or find context. The report of the researchers shows that reinforcement learning increased the result of Qwen3.5-9B on BrowseComp-Plus from 28.8% to 42.5%, with a 12% decrease in FLOPs.
Sources
Posted
