beep@piefed.worldM to Technology@piefed.worldEnglish · 14 hours agoNous Research unveils Arete-27B, a post-trained Qwen3.8-27B model that reinforces Western values, saying the underlying Chinese model's bias layer is shallownousresearch.comexternal-linkmessage-square3linkfedilinkarrow-up12arrow-down10 cross-posted to: hackernews@lemmy.bestiver.se
arrow-up12arrow-down1external-linkNous Research unveils Arete-27B, a post-trained Qwen3.8-27B model that reinforces Western values, saying the underlying Chinese model's bias layer is shallownousresearch.combeep@piefed.worldM to Technology@piefed.worldEnglish · 14 hours agomessage-square3linkfedilink cross-posted to: hackernews@lemmy.bestiver.se
minus-squareElectricVocalistlinkfedilinkarrow-up1·4 hours agoYou have clearly not read their methodology. The benchmark is telling the truth rather than the state propaganda version
You have clearly not read their methodology. The benchmark is telling the truth rather than the state propaganda version