\n","updatedAt":"2023-10-16T16:28:42.042Z","author":{"_id":"650e696ba0f2ffbeca687a3a","avatarUrl":"/avatars/d5365258828e5aecedc3cdf2768c4f60.svg","fullname":"PATHAN ASHRAF ","name":"Pathanashraf","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.29723411798477173},"editors":["Pathanashraf"],"editorAvatarUrls":["/avatars/d5365258828e5aecedc3cdf2768c4f60.svg"],"reactions":[],"isReport":false}},{"id":"652d64d6956d6a4244e0ffc0","author":{"_id":"650e696ba0f2ffbeca687a3a","avatarUrl":"/avatars/d5365258828e5aecedc3cdf2768c4f60.svg","fullname":"PATHAN ASHRAF ","name":"Pathanashraf","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false},"createdAt":"2023-10-16T16:29:10.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"Write a top 5points","html":"

Write a top 5points

\n","updatedAt":"2023-10-16T16:29:10.727Z","author":{"_id":"650e696ba0f2ffbeca687a3a","avatarUrl":"/avatars/d5365258828e5aecedc3cdf2768c4f60.svg","fullname":"PATHAN ASHRAF ","name":"Pathanashraf","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7698227167129517},"editors":["Pathanashraf"],"editorAvatarUrls":["/avatars/d5365258828e5aecedc3cdf2768c4f60.svg"],"reactions":[],"isReport":false}},{"id":"652d84ca956d6a4244e5e4b5","author":{"_id":"64660eae875b1a86a786c04e","avatarUrl":"/avatars/8b4900c358848aafd55194c12aa31bfe.svg","fullname":"Rohit Kumar","name":"Rohit-788","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false},"createdAt":"2023-10-16T18:45:30.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"\n\n","html":"

\n","updatedAt":"2023-10-16T18:45:30.764Z","author":{"_id":"64660eae875b1a86a786c04e","avatarUrl":"/avatars/8b4900c358848aafd55194c12aa31bfe.svg","fullname":"Rohit Kumar","name":"Rohit-788","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.3670727014541626},"editors":["Rohit-788"],"editorAvatarUrls":["/avatars/8b4900c358848aafd55194c12aa31bfe.svg"],"reactions":[],"isReport":false}},{"id":"652d84e546e3998e41e94d15","author":{"_id":"64660eae875b1a86a786c04e","avatarUrl":"/avatars/8b4900c358848aafd55194c12aa31bfe.svg","fullname":"Rohit Kumar","name":"Rohit-788","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false},"createdAt":"2023-10-16T18:45:57.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"start discussing about this paper\n","html":"

start discussing about this paper

\n","updatedAt":"2023-10-16T18:45:57.720Z","author":{"_id":"64660eae875b1a86a786c04e","avatarUrl":"/avatars/8b4900c358848aafd55194c12aa31bfe.svg","fullname":"Rohit Kumar","name":"Rohit-788","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8430055379867554},"editors":["Rohit-788"],"editorAvatarUrls":["/avatars/8b4900c358848aafd55194c12aa31bfe.svg"],"reactions":[],"isReport":false}},{"id":"652dadbdd838856129ef807a","author":{"_id":"6486638da4cf2081f20c40ec","avatarUrl":"/avatars/0bc16a7447cd71ac18828a678313bd83.svg","fullname":"Mike Young","name":"mikelabs","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":13,"isUserFollowing":false},"createdAt":"2023-10-16T21:40:13.000Z","type":"comment","data":{"edited":true,"hidden":false,"latest":{"raw":"Wtf is going on in these comments lol? Anyway, here's my summary...\n\nTables are everywhere - reports, databases, webpages. They neatly organize data for humans to parse. But despite strong language skills, AI still struggles with table comprehension.\n\nEven models like GPT-3 fail at basic tasks like finding where a missing value should go. This is because they're trained mostly on free-flowing text, not 2D tabular data. Unlike unstructured text, data in tables derives meaning from its structure and position!\n\nSo researchers at Microsoft tried \"table-tuning\" - extending training with synthesized table task cases. Tasks like \"impute missing value X\" or \"identify outliers in this table\". They did this using a corpus of real-world tables.\n\nThey also augmented the data more by paraphrasing, reordering rows/columns, chaining model responses, and more. This protects against overfitting.\n\nThe resulting Table-GPT models showed big improvements:\n\n* 25%+ better at unseen table tasks like missing value ID\n* Beat GPT-3 on 98% of test cases over 9 different table tasks\n* Stayed strong even after targeted downstream tuning\n\nTable-tuning seems a promising step toward AI that can handle tables. That would unlock automated analysis over the troves of valuable tabular data out there.\n\n**TLDR: Training models on a large and diverse dataset of synthesized table tasks significantly boosts their table skills.**\n\n[Full Summary is here.](https://notes.aimodels.fyi/table-gpt-table-tuned-gpt-for-diverse-table-tasks/)","html":"

Wtf is going on in these comments lol? Anyway, here's my summary...

\n

Tables are everywhere - reports, databases, webpages. They neatly organize data for humans to parse. But despite strong language skills, AI still struggles with table comprehension.

\n

Even models like GPT-3 fail at basic tasks like finding where a missing value should go. This is because they're trained mostly on free-flowing text, not 2D tabular data. Unlike unstructured text, data in tables derives meaning from its structure and position!

\n

So researchers at Microsoft tried \"table-tuning\" - extending training with synthesized table task cases. Tasks like \"impute missing value X\" or \"identify outliers in this table\". They did this using a corpus of real-world tables.

\n

They also augmented the data more by paraphrasing, reordering rows/columns, chaining model responses, and more. This protects against overfitting.

\n

The resulting Table-GPT models showed big improvements:

\n

\n
• 25%+ better at unseen table tasks like missing value ID
\n
• Beat GPT-3 on 98% of test cases over 9 different table tasks
\n
• Stayed strong even after targeted downstream tuning
\n

\n

Table-tuning seems a promising step toward AI that can handle tables. That would unlock automated analysis over the troves of valuable tabular data out there.

\n

TLDR: Training models on a large and diverse dataset of synthesized table tasks significantly boosts their table skills.

\n

Full Summary is here.

\n","updatedAt":"2023-10-16T21:44:29.235Z","author":{"_id":"6486638da4cf2081f20c40ec","avatarUrl":"/avatars/0bc16a7447cd71ac18828a678313bd83.svg","fullname":"Mike Young","name":"mikelabs","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":13,"isUserFollowing":false}},"numEdits":2,"identifiedLanguage":{"language":"en","probability":0.894295871257782},"editors":["mikelabs"],"editorAvatarUrls":["/avatars/0bc16a7447cd71ac18828a678313bd83.svg"],"reactions":[{"reaction":"🤗","users":["mikelabs","KrishnaKaasyap","amberberli","sciafri","jj97"],"count":5},{"reaction":"👍","users":["b08x","tdingman-scale","sciafri"],"count":3},{"reaction":"❤️","users":["sciafri","kkuhlman"],"count":2}],"isReport":false}},{"id":"652dd3db99687726aef79ffa","author":{"_id":"64aed48cfec303c461d06242","avatarUrl":"/avatars/236c771e6c5a25ef6ed5e1bc061e30b8.svg","fullname":"Krishna Kaasyap","name":"KrishnaKaasyap","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false},"createdAt":"2023-10-17T00:22:51.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"> Can you opensource your training datasets of the different table tasks?\n\nNot sure about that but here's a table dataset that is more than 800 billion tokens!\n\nhttps://huggingface.co/datasets/approximatelabs/tablib-v1-full","html":"

\n

Can you opensource your training datasets of the different table tasks?

\n

\n

Not sure about that but here's a table dataset that is more than 800 billion tokens!

\n

https://huggingface.co/datasets/approximatelabs/tablib-v1-full

\n","updatedAt":"2023-10-17T00:22:51.192Z","author":{"_id":"64aed48cfec303c461d06242","avatarUrl":"/avatars/236c771e6c5a25ef6ed5e1bc061e30b8.svg","fullname":"Krishna Kaasyap","name":"KrishnaKaasyap","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8816072940826416},"editors":["KrishnaKaasyap"],"editorAvatarUrls":["/avatars/236c771e6c5a25ef6ed5e1bc061e30b8.svg"],"reactions":[{"reaction":"👍","users":["jj97","LisaWang0306"],"count":2}],"isReport":false}},{"id":"6531189fba7ae55b6549179f","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":318,"isUserFollowing":false},"createdAt":"2023-10-19T11:53:03.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"This is an automated message from the [Librarian Bot](https://huggingface.co/librarian-bots). I found the following papers similar to this paper. \n\nThe following papers were recommended by the Semantic Scholar API \n\n* [Testing the Limits of Unified Sequence to Sequence LLM Pretraining on Diverse Table Data Tasks](https://huggingface.co/papers/2310.00789) (2023)\n* [LLM-augmented Preference Learning from Natural Language](https://huggingface.co/papers/2310.08523) (2023)\n* [Efficient Finetuning Large Language Models For Vietnamese Chatbot](https://huggingface.co/papers/2309.04646) (2023)\n* [A Systematic Evaluation of Large Language Models on Out-of-Distribution Logical Reasoning Tasks](https://huggingface.co/papers/2310.09430) (2023)\n* [Training Generative Question-Answering on Synthetic Data Obtained from an Instruct-tuned Model](https://huggingface.co/papers/2310.08072) (2023)\n\n Please give a thumbs up to this comment if you found it helpful!\n\n If you want recommendations for any Paper on Hugging Face checkout [this](https://huggingface.co/spaces/librarian-bots/recommend_similar_papers) Space","html":"

This is an automated message from the Librarian Bot. I found the following papers similar to this paper.

\n

The following papers were recommended by the Semantic Scholar API

\n

\n
• Testing the Limits of Unified Sequence to Sequence LLM Pretraining on Diverse Table Data Tasks (2023)
\n
• LLM-augmented Preference Learning from Natural Language (2023)
\n
• Efficient Finetuning Large Language Models For Vietnamese Chatbot (2023)
\n
• A Systematic Evaluation of Large Language Models on Out-of-Distribution Logical Reasoning Tasks (2023)
\n
• Training Generative Question-Answering on Synthetic Data Obtained from an Instruct-tuned Model (2023)
\n

\n

Please give a thumbs up to this comment if you found it helpful!

\n

If you want recommendations for any Paper on Hugging Face checkout this Space

\n","updatedAt":"2023-10-19T11:53:03.673Z","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":318,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7445301413536072},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[],"isReport":false}},{"id":"654ef7110aa8eba4c209cd31","author":{"_id":"6348f06b5b1a5329cc2a3923","avatarUrl":"/avatars/33212f35f87e7bba5294a5750d96ee0f.svg","fullname":"Shashi Prakash Tripathi","name":"ShishuTripathi","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false},"createdAt":"2023-11-11T03:37:53.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"I am still not very sure with GPT capabilities of tackling arithmetical calculation in any context let it be tables or time series data. Till now the output that I have seen is not even close.\n","html":"

I am still not very sure with GPT capabilities of tackling arithmetical calculation in any context let it be tables or time series data. Till now the output that I have seen is not even close.

\n","updatedAt":"2023-11-11T03:37:53.130Z","author":{"_id":"6348f06b5b1a5329cc2a3923","avatarUrl":"/avatars/33212f35f87e7bba5294a5750d96ee0f.svg","fullname":"Shashi Prakash Tripathi","name":"ShishuTripathi","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9469090104103088},"editors":["ShishuTripathi"],"editorAvatarUrls":["/avatars/33212f35f87e7bba5294a5750d96ee0f.svg"],"reactions":[],"isReport":false}},{"id":"655bfe8f948930b0fce7d10f","author":{"_id":"63f4afd0cfd7ba6e26f5d0c5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63f4afd0cfd7ba6e26f5d0c5/imFdETtKl94fpQefk0bon.jpeg","fullname":"Damian","name":"QuantumDamian","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false},"createdAt":"2023-11-21T00:49:19.000Z","type":"comment","data":{"edited":true,"hidden":false,"latest":{"raw":"Key conclusion from my review of v1 of this paper (published on 13th Oct 2023):\n\nThis paper offers good introduction to simple table-tuning tasks, however task T-3 (TQA) should be significantly improved before Table-GPT can be used commercially.\n\nKey points:\n•\tGeneric overview of results indicates very good results of table-tuning, however in my opinion tasks should be divided based on complexity to better understand value of table-tuning. Please see diagram below.\n\n\n\n•\tFor the most of easy tasks (all tasks except T-3) table-tuning offers great improvements in zero-shot comparing to vanilla models (209% improvement for GPT-3.5 and 119% for ChatGPT) \n•\tFor the most of easy tasks (all tasks except T-3) table-tuning offers good results in few-shot comparing to vanilla models (22% improvement for GPT-3.5 and 12% for ChatGPT)\n \n\n\n•\tT-3 TQA is the most complex task (and with biggest business demand) and for this tasks table-tuning offers very small improvements (1-2% for ChatGPT and 5-8% for GPT-3.5), which is probably not worth of the fine-tuning effort\n \n\n\nOpen questions:\n•\tDo you have plans to fine-tune GPT-4?\n•\tCan you share recommendations on improving T-3 (TQA)? Maybe including TQA tasks in training?\n•\tCan you include as well T-12 (NS) in tests?\n•\tCan you specify number of tokens used (both for training and test execution) for each task\n\nOther remarks:\n•\tMarkdown format increases performance of table-tuning by 3% comparing to CSV and by 5% comparing to JSON (table 5)\n•\tFor most of the tasks few-shot offers strong improvement over zero-shot for vanilla GPT 3.5 and ChatGPT (even without table-tuning).\n•\tTypos found in paper: \n-\tp.4 is “toke-by-token” should be “token-by-token”\n-\tp.6 is “few select table-tasks” should be “few selected table-tasks”\n-\tp.7 is “describes the row-augmentation task” should be “describes the column-augmentation task”","html":"

Key conclusion from my review of v1 of this paper (published on 13th Oct 2023):

\n

This paper offers good introduction to simple table-tuning tasks, however task T-3 (TQA) should be significantly improved before Table-GPT can be used commercially.

\n

Key points:
•\tGeneric overview of results indicates very good results of table-tuning, however in my opinion tasks should be divided based on complexity to better understand value of table-tuning. Please see diagram below.

\n

\n

•\tFor the most of easy tasks (all tasks except T-3) table-tuning offers great improvements in zero-shot comparing to vanilla models (209% improvement for GPT-3.5 and 119% for ChatGPT)
•\tFor the most of easy tasks (all tasks except T-3) table-tuning offers good results in few-shot comparing to vanilla models (22% improvement for GPT-3.5 and 12% for ChatGPT)

\n

\n

•\tT-3 TQA is the most complex task (and with biggest business demand) and for this tasks table-tuning offers very small improvements (1-2% for ChatGPT and 5-8% for GPT-3.5), which is probably not worth of the fine-tuning effort

\n

\n

Open questions:
•\tDo you have plans to fine-tune GPT-4?
•\tCan you share recommendations on improving T-3 (TQA)? Maybe including TQA tasks in training?
•\tCan you include as well T-12 (NS) in tests?
•\tCan you specify number of tokens used (both for training and test execution) for each task

\n

Other remarks:
•\tMarkdown format increases performance of table-tuning by 3% comparing to CSV and by 5% comparing to JSON (table 5)
•\tFor most of the tasks few-shot offers strong improvement over zero-shot for vanilla GPT 3.5 and ChatGPT (even without table-tuning).
•\tTypos found in paper:
-\tp.4 is “toke-by-token” should be “token-by-token”
-\tp.6 is “few select table-tasks” should be “few selected table-tasks”
-\tp.7 is “describes the row-augmentation task” should be “describes the column-augmentation task”

\n","updatedAt":"2023-11-21T01:20:37.140Z","author":{"_id":"63f4afd0cfd7ba6e26f5d0c5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63f4afd0cfd7ba6e26f5d0c5/imFdETtKl94fpQefk0bon.jpeg","fullname":"Damian","name":"QuantumDamian","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":3,"identifiedLanguage":{"language":"en","probability":0.8828281164169312},"editors":["QuantumDamian"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/63f4afd0cfd7ba6e26f5d0c5/imFdETtKl94fpQefk0bon.jpeg"],"reactions":[],"isReport":false}},{"id":"666557bf21aa69e386b12004","author":{"_id":"6186ddf6a7717cb375090c01","avatarUrl":"/avatars/716b6a7d1094c8036b2a8a7b9063e8aa.svg","fullname":"Julien BLANCHON","name":"blanchon","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":176,"isUserFollowing":false},"createdAt":"2024-06-09T07:20:31.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"# Revolutionizing AI: Table-GPT Enhances Language Models for Complex Table Tasks!\n\nhttps://cdn-uploads.huggingface.co/production/uploads/6186ddf6a7717cb375090c01/YBXfRJZdPO5YRXR-qpOr0.mp4 \n\n## Links 🔗:\n👉 Subscribe: https://www.youtube.com/@Arxflix\n👉 Twitter: https://x.com/arxflix\n👉 LMNT (Partner): https://lmnt.com/\n\n\nBy Arxflix\n","html":"

\n\t\n\t\t\n\t\n\t\n\t\tRevolutionizing AI: Table-GPT Enhances Language Models for Complex Table Tasks!\n\t\n

\n

\n\n

\n\t\n\t\t\n\t\n\t\n\t\tLinks 🔗:\n\t\n

\n

👉 Subscribe: https://www.youtube.com/@Arxflix
👉 Twitter: https://x.com/arxflix
👉 LMNT (Partner): https://lmnt.com/

\n

By Arxflix

\n","updatedAt":"2024-06-09T07:20:31.232Z","author":{"_id":"6186ddf6a7717cb375090c01","avatarUrl":"/avatars/716b6a7d1094c8036b2a8a7b9063e8aa.svg","fullname":"Julien BLANCHON","name":"blanchon","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":176,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.5163402557373047},"editors":["blanchon"],"editorAvatarUrls":["/avatars/716b6a7d1094c8036b2a8a7b9063e8aa.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2310.09263","authors":[{"_id":"652c9b55a2d97e682b62b270","name":"Peng Li","hidden":false},{"_id":"652c9b55a2d97e682b62b271","name":"Yeye He","hidden":false},{"_id":"652c9b55a2d97e682b62b272","name":"Dror Yashar","hidden":false},{"_id":"652c9b55a2d97e682b62b273","name":"Weiwei Cui","hidden":false},{"_id":"652c9b55a2d97e682b62b274","name":"Song Ge","hidden":false},{"_id":"652c9b55a2d97e682b62b275","name":"Haidong Zhang","hidden":false},{"_id":"652c9b55a2d97e682b62b276","name":"Danielle Rifinski Fainman","hidden":false},{"_id":"652c9b55a2d97e682b62b277","name":"Dongmei Zhang","hidden":false},{"_id":"652c9b55a2d97e682b62b278","name":"Surajit Chaudhuri","hidden":false}],"publishedAt":"2023-10-13T17:20:56.000Z","submittedOnDailyAt":"2023-10-16T00:39:25.588Z","title":"Table-GPT: Table-tuned GPT for Diverse Table Tasks","submittedOnDailyBy":{"_id":"60f1abe7544c2adfd699860c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674929746905-60f1abe7544c2adfd699860c.jpeg","isPro":false,"fullname":"AK","user":"akhaliq","type":"user"},"summary":"Language models, such as GPT-3.5 and ChatGPT, demonstrate remarkable\nabilities to follow diverse human instructions and perform a wide range of\ntasks. However, when probing language models using a range of basic\ntable-understanding tasks, we observe that today's language models are still\nsub-optimal in many table-related tasks, likely because they are pre-trained\npredominantly on one-dimensional natural-language texts, whereas\nrelational tables are two-dimensional objects.\n In this work, we propose a new \"table-tuning\" paradigm, where we\ncontinue to train/fine-tune language models like GPT-3.5 and ChatGPT, using\ndiverse table-tasks synthesized from real tables as training data, with the\ngoal of enhancing language models' ability to understand tables and perform\ntable tasks. We show that our resulting Table-GPT models demonstrate (1) better\ntable-understanding capabilities, by consistently outperforming the\nvanilla GPT-3.5 and ChatGPT, on a wide-range of table tasks, including holdout\nunseen tasks, and (2) strong generalizability, in its ability to respond\nto diverse human instructions to perform new table-tasks, in a manner similar\nto GPT-3.5 and ChatGPT.","upvotes":40,"discussionId":"652c9b55a2d97e682b62b284","ai_summary":"A new table-tuning paradigm improves language models' understanding and generalization in table-related tasks by fine-tuning them with synthesized table data, resulting in enhanced performance.","ai_keywords":["table-tuning","table-understanding","generalizability","Table-GPT","GPT-3.5","ChatGPT","table-tasks"]},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"61e9db1b7346f0a8ff5904e4","avatarUrl":"/avatars/ece6cabd401b68e3fc2743969a9c99f0.svg","isPro":false,"fullname":"Jan Zheng","user":"yawnxyz","type":"user"},{"_id":"6332e38ba652574b0cea2da4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6332e38ba652574b0cea2da4/YWVE8MRz7aCXLQ31CEhLT.png","isPro":true,"fullname":"Ismail Pelaseyed","user":"homanp","type":"user"},{"_id":"642e6cbb113b200fd7676bb1","avatarUrl":"/avatars/f1e3e2dc8bad30e88a5159732e4c5d0c.svg","isPro":false,"fullname":"fujingling","user":"FuJingLing","type":"user"},{"_id":"60e35a2de9edab19748c73e1","avatarUrl":"/avatars/0ca2be9455800f8a50072c277c77136b.svg","isPro":false,"fullname":"Mustafa Akben","user":"mustafaakben","type":"user"},{"_id":"64b9206d77ae61bcc80e858a","avatarUrl":"/avatars/f4a36b81ad43f13f3ac07f9889fa193d.svg","isPro":false,"fullname":"RitchieBlackmore","user":"RitchieBlackmoresRainbow","type":"user"},{"_id":"6453f69c13dba495b2df9c75","avatarUrl":"/avatars/c2e405370d263fe09a2c92ec36e0a08a.svg","isPro":false,"fullname":"setareh z","user":"setareh1","type":"user"},{"_id":"64d257481a6275df575ad17f","avatarUrl":"/avatars/154eabc6eb149f6e9a9b79a23f6e6b96.svg","isPro":false,"fullname":"Naethan Jacob","user":"naejac","type":"user"},{"_id":"5dd96eb166059660ed1ee413","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/5dd96eb166059660ed1ee413/NQtzmrDdbG0H8qkZvRyGk.jpeg","isPro":true,"fullname":"Julien Chaumond","user":"julien-c","type":"user"},{"_id":"6476c2cd592bb80eba2b7696","avatarUrl":"/avatars/3cd83b00a028af4bc2a10c78f57430d1.svg","isPro":false,"fullname":"ANUBHAV","user":"anubhav10mishra","type":"user"},{"_id":"63ac9130e3b217fb36ce7177","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63ac9130e3b217fb36ce7177/tjZ70f4RgTBrDzES_N8CQ.jpeg","isPro":false,"fullname":"Kaushal Powar","user":"kaushalpowar","type":"user"},{"_id":"64945856d8b51eea62559a1e","avatarUrl":"/avatars/8562a08609f635a1ca0b1964f477d59a.svg","isPro":false,"fullname":"Matt Barr","user":"mattbarr","type":"user"},{"_id":"63e625112d2c508de9f92547","avatarUrl":"/avatars/2deb4f9e605d0dc30a31538d5c9aa300.svg","isPro":false,"fullname":"Shyam Peri","user":"shyamperi","type":"user"}],"acceptLanguages":["*"],"dailyPaperRank":0}">