🪸NLP researcher, AI scientist at Deccan AI. Core interests: Computational Social Sciences, Conversational AI, Ai safety and Multilingualityhttps://bsky.app/profile/guneetsk.bsky.socialhttps://bsky.app/profile/guneetsk.bsky.social/post/3lc4dt54lgk2yLooking for mentors or collaborators. Do check it out #NLProc #llm [contains quote post or other embedded content]29 Nov 2024 19:45 +0000at://did:plc:un5ldyp3deaaukp3zrm3mlws/app.bsky.feed.post/3lc4dt54lgk2yhttps://bsky.app/profile/guneetsk.bsky.social/post/3lbz5c4gfu22u* Red Teaming and Hate Speech * I recently organized a survey which was successful in collecting the details about stereotypes and ideologies in the Indian Context. Based on the survey I designed a red teaming dataset in Hindi and English which has been able to break GPT4, Llama Family and Claude28 Nov 2024 13:10 +0000at://did:plc:un5ldyp3deaaukp3zrm3mlws/app.bsky.feed.post/3lbz5c4gfu22uhttps://bsky.app/profile/guneetsk.bsky.social/post/3lbpwgx3nza2xReposting since I guess I have better number of followers now 🤓 [contains quote post or other embedded content]24 Nov 2024 21:14 +0000at://did:plc:un5ldyp3deaaukp3zrm3mlws/app.bsky.feed.post/3lbpwgx3nza2xhttps://bsky.app/profile/guneetsk.bsky.social/post/3lbjzz3u3sk2wI am currently with a research group which has access to more than 100 in house annotators capable of designing human validated datasets, preferences and human in the loop experiments. Completed crazy Function Calling and Red Teaming exp with them. What else could I go for ?👀 #NLProc22 Nov 2024 13:02 +0000at://did:plc:un5ldyp3deaaukp3zrm3mlws/app.bsky.feed.post/3lbjzz3u3sk2whttps://bsky.app/profile/guneetsk.bsky.social/post/3lbfhe346k22cTrying to overfit a small language model into a knowledge base. Anyone with insights on such work #NLProc20 Nov 2024 17:17 +0000at://did:plc:un5ldyp3deaaukp3zrm3mlws/app.bsky.feed.post/3lbfhe346k22c