rajan agarwal @_rajanagarwal · 25 Nov 2025
in order to have research agents that can run for days, we need context compaction i used RL to have LLMs naturally learn their own 10x compression! Qwen learned to pack more info per token (ie use Mandarin tokens, prune text) read the technical blog: https://t.co/3pzRvt4zvr
202 943Views
891Likes
58Reposts
27Replies
14Quotes
953Bookmarks
Is that a lot?
7.02×vs this author's median28 902 views is typical
57Percentile for this authorof 7 recent posts
124×vs 10K–100K median1 637 views is typical
16.4× audienceReachviews ÷ followers
0.49%Engagement rateof viewers reacted