Experimental browser for the Atmosphere
This is a pure theory paper, but I found the process of working on it really clarified my thinking around computational vs. statistical benefits of interventions for RL with language models. Hopefully it will be a useful starting point toward a deeper understanding of how to explore efficiently. 17/
Mar 27, 2025, 5:28 PM
{ "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3llet5vhgo52c", "cid": "bafyreicmvkqw6zcvkfhbvwpl2lwcwtmqvzqrlk3fwqekcdlxzorjx2te6u", "value": { "text": "This is a pure theory paper, but I found the process of working on it really clarified my thinking around computational vs. statistical benefits of interventions for RL with language models. Hopefully it will be a useful starting point toward a deeper understanding of how to explore efficiently.\n17/", "$type": "app.bsky.feed.post", "langs": [ "en" ], "reply": { "root": { "cid": "bafyreidzv7o2gllewwfqqi4ixremjopxmuyntc5gstydqektrxkcjni4pi", "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3llet5p66ac2c" }, "parent": { "cid": "bafyreihjh4w6x5gzlmcl2tm6pjbjrt56pbuihknvgzyipqteocjbbscp2u", "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3llet5usuhn2c" } }, "createdAt": "2025-03-27T17:28:13.786Z" } }