フラジークHladik・ラジムRadim

@[email protected]
178 Followers
342 Following
173 Posts

#sciencestudies | #researchfunding | #scholarlycommunication | #digitalhumanities | #computationalsocialscience | #scienceofscience | #rstats | #VedaVyzkum

Czech Academy of Sciences

フラジークHladik・ラジムRadimNov 23, 2024
CRIS_SciencesPoNov 22, 2024
Mapping knowledge: Topic analysis of science locates researchers in disciplinary landscape. Science is split by Culture–Nature, Life–Non-life, and Materials–Methods dichotomies. Epistemological position-takings reveal social patterns among knowledge producers. Paper by Yann Renisio & Radim Hladík https://www.sciencedirect.com/science/article/pii/S0304422X24000895 or Preprint https://osf.io/preprints/socarxiv/94jd5
Mapping knowledge: Topic analysis of science locates researchers in disciplinary landscape

The study presents a new approach for constructing an epistemological coordinate system that locates individual researchers within the disciplinary la…

フラジークHladik・ラジムRadimOct 28, 2024
Michel NivardOct 26, 2024
Extremely janky code in various places, but I think I got almost all of the R visualisation/interactive widget code to work within Obsidian, and am close to an implementation of R that does 90% of what you'd need for personal note-taking. #rstats #Obsidian
フラジークHladik・ラジムRadimOct 14, 2024
Robert SeyfertOct 14, 2024

It seems that the Noble Memorial Prize in Economic Sciences has gone to researchers who have worked on the problem of how to colonise a country better.

https://www.nobelprize.org/prizes/economic-sciences/2024/press-release/

The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2024

The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2024 was awarded jointly to Daron Acemoglu, Simon Henry Roberts Johnson and James A. Robinson "for studies of how institutions are formed and affect prosperity"

NobelPrize.org
フラジークHladik・ラジムRadimOct 9, 2024
Gabor CsardiOct 9, 2024

Nice post by Stephen Turner about using Parquet files instead of CSV, with nanoparquet and/or duckdb: https://blog.stephenturner.us/p/use-nanoparquet-instead-of-readr-csv

#rstats

Use nanoparquet instead of readr/CSV

Parquet is interoperable between Python and R, fast to read+write, works well with databases, and stores complex data types (e.g., tibble listcols). Use it instead of CSV. Many pros, few (no?) cons.

Paired Ends
フラジークHladik・ラジムRadimOct 9, 2024
CWTS | Leiden UniversityOct 9, 2024

📊 Today, CWTS has launched the Leiden Ranking Open Edition 2024! 🎉

As in the first release in January, the new version is completely based on @OpenAlex data, made possible by Our Research, and provides an open and transparent alternative to traditional university rankings.

https://open.leidenranking.com/

CWTS Leiden Ranking Open Edition

The CWTS Leiden Ranking Open Edition offers important insights into the scientific performance of over 1500 major universities worldwide. Select your preferred indicators, generate results, and explore the performance of universities.

CWTS Leiden Ranking Open Edition
フラジークHladik・ラジムRadimSep 24, 2024
Inverse Complexity LabSep 23, 2024

Hello world! We're a research group led by Tiago Peixoto (@tiago) at the newly founded IT:U, Linz, Austria.

We focus on inverse problems in network science and complex systems.

Follow this account to get scientific updates!

https://skewed.de/lab

#networkscience #complexsystems

Inverse Complexity Lab – Tiago P. Peixoto

Inverse Complexity Lab

skewed.de
フラジークHladik・ラジムRadimAug 28, 2024
Ulrike HahnAug 27, 2024

new blog post on whether LLMs really reason, think, summarise etc.

as always, all comments welcome

https://write.as/ulrikehahn/do-llms-really-reason-understand-think-summarise

#LLM #LLMs #AI #generativeAI @cogsci @philosophy

Do LLMs REALLY reason, understand, think, summarise...?

One way in which discussions of AI capabilities are unsatisfying is that they often descend into what feels like argument about words or ...

UlrikeHahn
フラジークHladik・ラジムRadimAug 25, 2024
Olivier GodechotAug 24, 2024
It’s there ! https://www.journals.uchicago.edu/doi/10.1086/731603 . With a little work (9 years), some data points (1 billion +) from a handful of OECD countries (12) during a couple of years (30) and a few coauthors (28 :
1/n
フラジークHladik・ラジムRadimAug 19, 2024
YY AhnAug 19, 2024

It's intriguing that there are neurons in LLMs that play specific functional roles such as regulating confidence (https://arxiv.org/abs/2406.16254) and that they may consistently emerge across different realizations of the LLMs (https://arxiv.org/abs/2401.12181).

Any biological analogues in human brain?

Confidence Regulation Neurons in Language Models

Despite their widespread use, the mechanisms by which large language models (LLMs) represent and regulate uncertainty in next-token predictions remain largely unexplored. This study investigates two critical components believed to influence this uncertainty: the recently discovered entropy neurons and a new set of components that we term token frequency neurons. Entropy neurons are characterized by an unusually high weight norm and influence the final layer normalization (LayerNorm) scale to effectively scale down the logits. Our work shows that entropy neurons operate by writing onto an unembedding null space, allowing them to impact the residual stream norm with minimal direct effect on the logits themselves. We observe the presence of entropy neurons across a range of models, up to 7 billion parameters. On the other hand, token frequency neurons, which we discover and describe here for the first time, boost or suppress each token's logit proportionally to its log frequency, thereby shifting the output distribution towards or away from the unigram distribution. Finally, we present a detailed case study where entropy neurons actively manage confidence in the setting of induction, i.e. detecting and continuing repeated subsequences.

arXiv.org
フラジークHladik・ラジムRadimAug 18, 2024
McDonald_69Aug 18, 2024

This ....

#UkraineWar