Open ronentk opened 10 months ago
Sub-issue from #24
https://twitter.com/victorveitch/status/1722300572554969090
Tweet contains arxiv link:
Scrape tweet fails to extract it:
{'conversationID': '1722300572554969090', 'date': 'Wed Nov 08 17:10:29 +0000 2023', 'date_epoch': 1699463429, 'hashtags': [], 'likes': 1465, 'mediaURLs': ['https://pbs.twimg.com/media/F-bW6ZnXsAABCu9.png', 'https://pbs.twimg.com/media/F-bXIlSXgAEl61P.jpg'], 'media_extended': [{'altText': None, 'size': {'height': 671, 'width': 1645}, 'thumbnail_url': 'https://pbs.twimg.com/media/F-bW6ZnXsAABCu9.png', 'type': 'image', 'url': 'https://pbs.twimg.com/media/F-bW6ZnXsAABCu9.png'}, {'altText': None, 'size': {'height': 676, 'width': 1092}, 'thumbnail_url': 'https://pbs.twimg.com/media/F-bXIlSXgAEl61P.jpg', 'type': 'image', 'url': 'https://pbs.twimg.com/media/F-bXIlSXgAEl61P.jpg'}], 'possibly_sensitive': False, 'qrtURL': None, 'replies': 18, 'retweets': 224, 'text': 'There\'s an idea that LLMs encode high-level concepts linearly in representation space, and that these can be understood using geometric operations (e.g., cosine similarity) \n\nBut: What does "linear" even mean? And, why would (Euclidean) geometry encode meaning?… https://t.co/8WXQVI4WBb', 'tweetID': '1722300572554969090', 'tweetURL': 'https://twitter.com/victorveitch/status/1722300572554969090', 'user_name': 'Victor Veitch', 'user_screen_name': 'victorveitch'}
Sub-issue from #24
Tweet contains arxiv link:
Scrape tweet fails to extract it: