Experimental browser for the Atmosphere
Standard training techniques like good data curation, SFT -> DPO, work reasonably well, and the pass/fail nature of guardrail adherence enables the use of tricks like classifier-free guidance/contrastive decoding to further improve performance
Feb 19, 2025, 6:06 AM
{ "uri": "at://did:plc:aesb2dqqqmodtvf24fs75y4r/app.bsky.feed.post/3lij4bn4zxs2i", "cid": "bafyreia4qvagzqmoevt3vfboj4qluonqwzepvspb7obixpq6yjmslgc5lm", "value": { "text": "Standard training techniques like good data curation, SFT -> DPO, work reasonably well, and the pass/fail nature of guardrail adherence enables the use of tricks like classifier-free guidance/contrastive decoding to further improve performance", "$type": "app.bsky.feed.post", "embed": { "$type": "app.bsky.embed.images", "images": [ { "alt": "", "image": { "$type": "blob", "ref": { "$link": "bafkreicn24rgc5eitf6hgjl7j23eltqrsft7ud6egiswxzwr7uhc6trina" }, "mimeType": "image/jpeg", "size": 191405 }, "aspectRatio": { "width": 2000, "height": 457 } } ] }, "langs": [ "en" ], "reply": { "root": { "cid": "bafyreigosybciamvspmonowrrdh6msea73k3hfsexshct6xeyf5f74tvc4", "uri": "at://did:plc:aesb2dqqqmodtvf24fs75y4r/app.bsky.feed.post/3lij4bkvgxc2i" }, "parent": { "cid": "bafyreidhkjgnm4vcevolljeb25ac6zdu5iu2iw2cgbos7kihk2bz73gyw4", "uri": "at://did:plc:aesb2dqqqmodtvf24fs75y4r/app.bsky.feed.post/3lij4blvu2s2i" } }, "createdAt": "2025-02-19T06:06:31.255Z" } }