Tim Baumgärtner
@timbmg.bsky.social
👨💻 NLP PhD Student @ukplab.bsky.social
💡 TIL @overleaf.com is basically a git repo.
In my research workflow, I directly added it as submodule to my code repo. Now I can produce figures and tables, and have them magically uploaded to Overleaf just by pushing the repo.
No more renaming, keeping versions straight, and manual uploading 😇
In my research workflow, I directly added it as submodule to my code repo. Now I can produce figures and tables, and have them magically uploaded to Overleaf just by pushing the repo.
No more renaming, keeping versions straight, and manual uploading 😇
November 10, 2025 at 10:44 AM
💡 TIL @overleaf.com is basically a git repo.
In my research workflow, I directly added it as submodule to my code repo. Now I can produce figures and tables, and have them magically uploaded to Overleaf just by pushing the repo.
No more renaming, keeping versions straight, and manual uploading 😇
In my research workflow, I directly added it as submodule to my code repo. Now I can produce figures and tables, and have them magically uploaded to Overleaf just by pushing the repo.
No more renaming, keeping versions straight, and manual uploading 😇
💡 TIL, it's super easy to fetch data from Google Sheets into Pandas. Makes it really convenient to annotate some data.
Previously, I was always downloading CSVs, losing track of file versions, and loading and merging them sluggishly in Python.
👉 find the code here: gist.github.com/timbmg/6c2d6...
Previously, I was always downloading CSVs, losing track of file versions, and loading and merging them sluggishly in Python.
👉 find the code here: gist.github.com/timbmg/6c2d6...
November 4, 2025 at 4:00 PM
💡 TIL, it's super easy to fetch data from Google Sheets into Pandas. Makes it really convenient to annotate some data.
Previously, I was always downloading CSVs, losing track of file versions, and loading and merging them sluggishly in Python.
👉 find the code here: gist.github.com/timbmg/6c2d6...
Previously, I was always downloading CSVs, losing track of file versions, and loading and merging them sluggishly in Python.
👉 find the code here: gist.github.com/timbmg/6c2d6...
Reposted by Tim Baumgärtner
🔍 𝗪𝗮𝗻𝘁 𝘁𝗼 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗲 𝗺𝗼𝗱𝗲𝗹𝘀 𝗼𝗻 𝘀𝗰𝗶𝗲𝗻𝘁𝗶𝗳𝗶𝗰 𝗤𝗔, 𝗯𝘂𝘁 𝘆𝗼𝘂𝗿 𝗱𝗮𝘁𝗮𝘀𝗲𝘁 𝗹𝗮𝗰𝗸𝘀 𝗿𝗲𝗮𝗹-𝘄𝗼𝗿𝗹𝗱 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀 𝗮𝘀𝗸𝗲𝗱 𝗯𝘆 𝗲𝘅𝗽𝗲𝗿𝘁𝘀?
🚀 PeerQA is the solution: a dataset with questions from peer reviews and answers from the original authors. (1/🧵)
#NLProc
🚀 PeerQA is the solution: a dataset with questions from peer reviews and answers from the original authors. (1/🧵)
#NLProc
April 25, 2025 at 7:46 AM
🔍 𝗪𝗮𝗻𝘁 𝘁𝗼 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗲 𝗺𝗼𝗱𝗲𝗹𝘀 𝗼𝗻 𝘀𝗰𝗶𝗲𝗻𝘁𝗶𝗳𝗶𝗰 𝗤𝗔, 𝗯𝘂𝘁 𝘆𝗼𝘂𝗿 𝗱𝗮𝘁𝗮𝘀𝗲𝘁 𝗹𝗮𝗰𝗸𝘀 𝗿𝗲𝗮𝗹-𝘄𝗼𝗿𝗹𝗱 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀 𝗮𝘀𝗸𝗲𝗱 𝗯𝘆 𝗲𝘅𝗽𝗲𝗿𝘁𝘀?
🚀 PeerQA is the solution: a dataset with questions from peer reviews and answers from the original authors. (1/🧵)
#NLProc
🚀 PeerQA is the solution: a dataset with questions from peer reviews and answers from the original authors. (1/🧵)
#NLProc
Reposted by Tim Baumgärtner
🚨🚨 New preprint 🚨🚨
Ever wonder whether verbalized CoTs correspond to the internal reasoning process of the model?
We propose a novel parametric faithfulness approach, which erases information contained in CoT steps from the model parameters to assess CoT faithfulness.
arxiv.org/abs/2502.14829
Ever wonder whether verbalized CoTs correspond to the internal reasoning process of the model?
We propose a novel parametric faithfulness approach, which erases information contained in CoT steps from the model parameters to assess CoT faithfulness.
arxiv.org/abs/2502.14829
Measuring Faithfulness of Chains of Thought by Unlearning Reasoning Steps
When prompted to think step-by-step, language models (LMs) produce a chain of thought (CoT), a sequence of reasoning steps that the model supposedly used to produce its prediction. However, despite mu...
arxiv.org
February 21, 2025 at 12:43 PM
🚨🚨 New preprint 🚨🚨
Ever wonder whether verbalized CoTs correspond to the internal reasoning process of the model?
We propose a novel parametric faithfulness approach, which erases information contained in CoT steps from the model parameters to assess CoT faithfulness.
arxiv.org/abs/2502.14829
Ever wonder whether verbalized CoTs correspond to the internal reasoning process of the model?
We propose a novel parametric faithfulness approach, which erases information contained in CoT steps from the model parameters to assess CoT faithfulness.
arxiv.org/abs/2502.14829
Reposted by Tim Baumgärtner
𝗙𝗮𝗰𝘁-𝗖𝗵𝗲𝗰𝗸𝗶𝗻𝗴 𝗶𝗻 𝘁𝗵𝗲 𝗔𝗴𝗲 𝗼𝗳 𝗔𝗜 – 𝗔 𝗧𝗮𝗹𝗸 𝗯𝘆 𝗜𝗿𝘆𝗻𝗮 𝗚𝘂𝗿𝗲𝘃𝘆𝗰𝗵 @𝗔𝗜 𝗳𝗼𝗿 𝗚𝗼𝗼𝗱
Misinformation is a new weapon disrupting public debates, scientific discussions, and political decisions. How can we identify and counter misleading content?
(1/🧵)
Misinformation is a new weapon disrupting public debates, scientific discussions, and political decisions. How can we identify and counter misleading content?
(1/🧵)
Towards real-world fact-checking with large language models
Misinformation poses a growing threat to our society. It has a severe impact on public health by promoting fake cures fear and distrust. Current research
aiforgood.itu.int
February 18, 2025 at 8:27 AM
𝗙𝗮𝗰𝘁-𝗖𝗵𝗲𝗰𝗸𝗶𝗻𝗴 𝗶𝗻 𝘁𝗵𝗲 𝗔𝗴𝗲 𝗼𝗳 𝗔𝗜 – 𝗔 𝗧𝗮𝗹𝗸 𝗯𝘆 𝗜𝗿𝘆𝗻𝗮 𝗚𝘂𝗿𝗲𝘃𝘆𝗰𝗵 @𝗔𝗜 𝗳𝗼𝗿 𝗚𝗼𝗼𝗱
Misinformation is a new weapon disrupting public debates, scientific discussions, and political decisions. How can we identify and counter misleading content?
(1/🧵)
Misinformation is a new weapon disrupting public debates, scientific discussions, and political decisions. How can we identify and counter misleading content?
(1/🧵)
Reposted by Tim Baumgärtner
🤔 An Energy Star for AI? Introducing AI Energy Score: First-ever rating system comparing 166 AI models' energy consumption!
From LLaMa to Gemma, get transparent ⭐️1-5 efficiency ratings.
Incredible work led by @sashamtl.bsky.social
huggingface.co/blog/sasha/a...
From LLaMa to Gemma, get transparent ⭐️1-5 efficiency ratings.
Incredible work led by @sashamtl.bsky.social
huggingface.co/blog/sasha/a...
February 11, 2025 at 9:44 AM
🤔 An Energy Star for AI? Introducing AI Energy Score: First-ever rating system comparing 166 AI models' energy consumption!
From LLaMa to Gemma, get transparent ⭐️1-5 efficiency ratings.
Incredible work led by @sashamtl.bsky.social
huggingface.co/blog/sasha/a...
From LLaMa to Gemma, get transparent ⭐️1-5 efficiency ratings.
Incredible work led by @sashamtl.bsky.social
huggingface.co/blog/sasha/a...
Excited to share that our Paper "PeerQA: A Scientific Question Answering Dataset from Peer Reviews" as been accepted to #NAACL2025 Looking forward to presenting it in Albuquerque 🏜️!
We have +1 for #NAACL:
»PeerQA: A Scientific Question Answering Dataset from Peer Reviews« by Tim Baumgärtner (@timbmg.bsky.social), Ted Briscoe, Iryna Gurevych (@igurevych.bsky.social)
»PeerQA: A Scientific Question Answering Dataset from Peer Reviews« by Tim Baumgärtner (@timbmg.bsky.social), Ted Briscoe, Iryna Gurevych (@igurevych.bsky.social)
January 27, 2025 at 2:13 PM
Excited to share that our Paper "PeerQA: A Scientific Question Answering Dataset from Peer Reviews" as been accepted to #NAACL2025 Looking forward to presenting it in Albuquerque 🏜️!
Reposted by Tim Baumgärtner
What do YOU mean by "intelligence", and does ChatGPT fit your definition?
We collected the major criteria used in CogSci and other fields, and designed a survey to find out!
Access link: www.survey-xact.dk/collect
Code: 4S7V-SN4M-S536
Time: 5-10 mins
We collected the major criteria used in CogSci and other fields, and designed a survey to find out!
Access link: www.survey-xact.dk/collect
Code: 4S7V-SN4M-S536
Time: 5-10 mins
Perspectives on Intelligence: Community Survey
Research survey exploring how NLP/ML/CogSci researchers define and use the concept of intelligence.
bertramhojer.github.io
December 4, 2024 at 7:48 AM
What do YOU mean by "intelligence", and does ChatGPT fit your definition?
We collected the major criteria used in CogSci and other fields, and designed a survey to find out!
Access link: www.survey-xact.dk/collect
Code: 4S7V-SN4M-S536
Time: 5-10 mins
We collected the major criteria used in CogSci and other fields, and designed a survey to find out!
Access link: www.survey-xact.dk/collect
Code: 4S7V-SN4M-S536
Time: 5-10 mins