Sign in

julianpollmann.bsky.social

@julianpollmann.bsky.social
33 followers 80 following 16 posts

bike maniac & research software engineer at HSD, @zdd-hsd.bsky.social Mostly #NLP, #AI, #SoftwareEngineering

PostsRepliesMedia
Reposted by @julianpollmann.bsky.social
Florian Huber @me-datapoint.bsky.social · 30/06/2026
Our fingerprint analysis work has now been published 🎉 --> link.springer.com/article/10.1... Also check out our small helper library #chemap --> github.com/matchms/chemap (work together with @julianpollmann.bsky.social) #cheminformatics #openscience #Python
link.springer.com
Count your bits: fingerprint benchmarking to assess broad chemical space representation - Journal of Cheminformatics
Quantifying molecular similarity is a cornerstone of cheminformatics, underpinning applications from virtual screening and nearest-neighbor search to chemical space visualization and the evaluation of...
142
julianpollmann.bsky.social @julianpollmann.bsky.social · 19/06/2026
Thanks Canyon, the only Domain where I could escape the AI Hype was riding my Bike: now AI ist integrated into a Road Bike. www.rennrad-news.de/news/canyon-...
rennrad-news.de
Canyon stellt Rennrad der Zukunft mit KI und Rundum-Radar vor
Canyons neues KI-Rennrad der Zukunft erkennt Gefahren, bevor sie entstehen, und könnte schwere Unfälle verhindern.
000
Reposted by @julianpollmann.bsky.social
Vicki @vickiboykis.com · 19/05/2026
New post: Tagging my blog with BERTopic and LLMs. I suspect as token costs increase, we'll see more blended traditional ML/LLM systems. And, in general, they work really well! vickiboykis.com/2026/05/18/t...
vickiboykis.com
Tagging my blog posts with BERTopic and LLMs
LLMs mean you still need a human in the loop, but in a different part
4549
julianpollmann.bsky.social @julianpollmann.bsky.social · 13/04/2026
Really looking forward to #PyConDe & #PyData 2026 edition! Some open Missions: ☑️ Soak in everything regarding Knowledge Graphs ☑️ Solve open mysteries in Python Multiprocessing ☑️ Meet new people and old friends ☑️ get a feel for the hot topics (except obvious ones) get there (partially) & back by bike
000
Reposted by @julianpollmann.bsky.social
Florian Huber @me-datapoint.bsky.social · 27/02/2026
Just updated our preprint on benchmarking molecular fingerprints! --> www.biorxiv.org/content/10.1... Some key points - Count fingerprints should be the default (not binary!) - Unfolded fingerprints are often worth it. - Larger radius Morgan or FCFP fingerprints are good first bets.
MST graph comparing the top-10 ranking overlap between many commonly used fingerprint types and variants.Four UMAP visualizations of chemical space based on 700k fingerprints on the biostructures dataset using various molecular fingerprint types.
172
Reposted by @julianpollmann.bsky.social
Florian Huber @me-datapoint.bsky.social · 27/02/2026
Work done with @julianpollmann.bsky.social at @zdd-hsd.bsky.social Code: - Central functionalities are now pip installable --> github.com/matchms/chemap - Notebooks for experiments --> github.com/florian-huber/molecular_fingerprint_comparisons #openscience #opensource #cheminformatics #python
github.com
GitHub - matchms/chemap: Library for computing molecular fingerprint based similarities as well as dimensionality-reduction-based chemical space visualizations.
Library for computing molecular fingerprint based similarities as well as dimensionality-reduction-based chemical space visualizations. - matchms/chemap
131
Reposted by @julianpollmann.bsky.social
The Guardian @theguardian.com · 12/11/2025
Haftbefehl shows that Germany loves art born from alienation – just not the people who create it
theguardian.com
Haftbefehl shows that Germany loves art born from alienation – just not the people who create it
A hit Netflix documentary about Germany’s favourite rapper demonstrates how popular the aesthetics of migrant life are – just as politicians debate how to remove it from inner cities
14411
Reposted by @julianpollmann.bsky.social
Florian Huber @me-datapoint.bsky.social · 06/10/2025
New #matchms release (0.31)🚀 With functionalities that were on our TODO list for a looooong time: Flash Entropy and BLINK scores! The new "FlashSimilarity" allows computing modified cosine, spectral entropy etc., about 100x faster (or more if you use Linux). #Python #opensource #massspec
162
Reposted by @julianpollmann.bsky.social
Florian Huber @me-datapoint.bsky.social · 23/06/2025
4/4 We also highlight options for count fingerprints, such as log-counts and IDF weighted counts. The latter can be used to adjust the bit importance to a dataset of your choice. An example use-case are chemical space visualizations. Preprint: www.biorxiv.org/content/10.1...
Chemical Space Visualizations using UMAP and various molecular fingerprints.
021
Reposted by @julianpollmann.bsky.social
Florian Huber @me-datapoint.bsky.social · 23/06/2025
3/4 A huge issue is bit collisions. Fingerprints with a high bit occupation (RDKit, MAP4) often lead to (1) arbitrary misinterpretations, (2) shifts to high Tanimoto scores, (3) very different handling of small and large molecules. --> Consider using sparse fingerprints! --> Morgan >> MAP4 / RDKit
121
Reposted by @julianpollmann.bsky.social
Florian Huber @me-datapoint.bsky.social · 23/06/2025
2/4 We focused on weaknesses of the fingerprints. Many show frequent duplicates, so same fingerprint for different compounds. Most problematic: this can include *very* different compounds ending up with identical fingerprints. - MAP4 >> Morgan-type >> daylight - count >> binary #cheminformatics
Benchmarking plot on fingerprint duplications.
101
Reposted by @julianpollmann.bsky.social
Florian Huber @me-datapoint.bsky.social · 23/06/2025
New preprint out! 1/4 @julianpollmann.bsky.social and I went down several rabbit holes to assess some commonly used molecular fingerprints. Bottom line: For large datasets, make an effort to select suitable settings. "We used Tanimoto" is not good enough. --> www.biorxiv.org/content/10.1...
Sketch of count/binary fingerprints and weighing options.
151
Reposted by @julianpollmann.bsky.social
hagen terschüren @ausnahmsweise.net · 28/05/2025
was @tante.cc sagt
Libertäre kennen 
• den Preis von allem 
• den Wert von gar nichts 
• das Schutzalter in jeder Juristiktion
2276
Reposted by @julianpollmann.bsky.social
Marina Weisband @afelia.bsky.social · 12/05/2025
Wir dachten, KI würde uns irgendwann durch Terminator-Roboter versklaven. Aber wir werden durch Bequemlichkeit versklavt.
41857104
julianpollmann.bsky.social @julianpollmann.bsky.social · 23/04/2025
Hello @pyconde.bsky.social Curious to listen to some interesting Talks and meet new people 😀
010
Reposted by @julianpollmann.bsky.social
Nathan Lambert @natolambert.bsky.social · 16/04/2025
First draft online version of The RLHF Book is DONE. Recently I've been creating the advanced discussion chapters on everything from Constitutional AI to evaluation and character training, but I also sneak in consistent improvements to the RL specific chapter. rlhfbook.com
212218
Reposted by @julianpollmann.bsky.social
Katharina Nocun @kattascha.bsky.social · 10/04/2025
„Bei der #SPD haben noch die Mitglieder die Chance, die Notbremse zu ziehen und den Abbau wichtiger Grundrechte zu verhindern. Wir appellieren daher an die Sozialdemokraten: Stimmt dieser Überwachungsliste nicht zu!“ 🤝👏 @ccc.de
ccc.de
CCC | CCC fordert Notbremse für den Überwachungskatalog im Koalitionsvertrag
Der Chaos Computer Club ist eine galaktische Gemeinschaft von Lebewesen für Informationsfreiheit und Technikfolgenabschätzung.
24618247
Reposted by @julianpollmann.bsky.social
Hilde Kuehne @hildekuehne.bsky.social · 05/04/2025
I think this is an important point that we missed the last 10y. We need to call out fiction as fiction and fighting for reality (can be boring, but it’s real). Reality is, there is no AGI, just language models. If you think they are giving you some truth, the stupidity is on you, not the LLM.
2295
Reposted by @julianpollmann.bsky.social
PyCon DE & PyData @pyconde.bsky.social · 05/04/2025
👨‍🏫 Can you help spread the word? Our community supports future #AI professionals like engineers and data analysts. With backing from PySV and Pioneers Hub, we provide students free virtual passes to Germany's top AI conference. Learn more at 2025.pycon.de/blog/free-re....
2025.pycon.de
Enabling the Next Generation in AI & Open Source: Free Remote Tickets for Students at PyCon DE & PyData
Join PyCon DE & PyData 2025 in Darmstadt (Frankfurt), April 23-25! Germany’s largest Python and Data Science conference with talks, workshops, and community events like DjangoGirls and sprints. Be…
023
julianpollmann.bsky.social @julianpollmann.bsky.social · 03/04/2025
Nice keynote by @tante.cc for the opening of @zdd-hsd.bsky.social #HSD
021
Reposted by @julianpollmann.bsky.social
Zentrum für Digitalisierung und Digitalität (ZDD) @zdd-hsd.bsky.social · 02/04/2025
Die Vorbereitungen für die Einweihung des #ZDD laufen auf Hochtouren! Wir freuen uns auf morgen.
ZDD Foyer bereit für die feierliche Eröffnung
022
Reposted by @julianpollmann.bsky.social
Stanford HAI @stanfordhai.bsky.social · 28/03/2025
Stanford scholars introduced an open-source AI agent that learns how to navigate websites by mimicking childhood learning – an approach that could lead to more efficient, transparent, and privacy-conscious AI: hai.stanford.edu/news/an-open... @chrmanning.bsky.social @shikharmurty.bsky.social
hai.stanford.edu
An Open-Source AI Agent for Doing Tasks on the Web | Stanford HAI
NNetNav learns how to navigate websites by mimicking childhood learning through exploration.
1208
Reposted by @julianpollmann.bsky.social
Amrei Bahr @amreibahr.bsky.social · 28/03/2025
#IchBinHanna & der 🐘 senden 1 Botschaft in die Koalitionsverhandlungen! #WissZeitVG @drkeichhorn.bsky.social @kubon.bsky.social
616344
julianpollmann.bsky.social @julianpollmann.bsky.social · 24/03/2025
Interesting paper on Human Cognition in the Age of #AI or how not to brainrot arxiv.org/html/2502.12...
arxiv.org
Protecting Human Cognition in the Age of AI
000
Reposted by @julianpollmann.bsky.social
Nils Petras @nilspetras.bsky.social · 24/03/2025
Die amerikanische Forschung ist auch für Deutschland wichtig. Das ist ja kein Wettbewerb, den wir gewinnen wollen. Wenn es anderen schlechter geht, wirkt sich das auch auf uns und unseren gesellschaftlichen Fortschritt aus.
2154
Reposted by @julianpollmann.bsky.social
tante @tante.cc · 21/03/2025
It is super weird how little reception David Golumbia's last book "Cyberlibertarianism" has had. I mean I know why, he dropped a bunch of uncomfortable truth bombs there but it's still weird how little that book is talked about. It is probably the most important book about tech in the last decade.
1224047
Reposted by @julianpollmann.bsky.social
John Skiles Skinner @skiles.blue · 21/03/2025
I helped build a government AI system. DOGE fired me, rolled the AI out to the whole agency, and implied the AI can do my job and the jobs of the others they've fired. It can't. But, what DOGE accidentally revealed about themselves in the process is fascinating. 🧵
191103884334
Reposted by @julianpollmann.bsky.social
Hilary Mason @hilarymason.com · 22/03/2025
A friend pointed out that the quasi-religious way so many tech guys talk about AI is in part because they don't want to grapple with the guilt of what they do; AI is a God, AI will fix it. No need to be responsible for what you put in the world. I can't stop thinking about that.
518646
Reposted by @julianpollmann.bsky.social
Katharina Nocun @kattascha.bsky.social · 20/03/2025
Heftig. Laut Datenbank sind alle meine 4 Bücher - 3 davon zusammen mit @pialamberty.bsky.social verfasst - im Datensatz enthalten.
Screenshot - Liste der Buchtitel
2119659
Reposted by @julianpollmann.bsky.social
ACL 2027 @aclmeeting.bsky.social · 18/03/2025
📢 #ACL2025NLP Student Research Workshop CFP is out 🙌 👉 2025.aclweb.org/calls/studen... Send your paper by ⏰ May 18th, 2025 🚀 #nlproc
2025.aclweb.org
Student Research Workshop
ACL 2025 Student Research Workshop.
062
Reposted by @julianpollmann.bsky.social
Hilde Kuehne @hildekuehne.bsky.social · 18/03/2025
Just as an update ... 🚀 Call for Papers – CVPR 3rd Workshop on Multi-Modal Foundation Models (MMFM) @cvprconference.bsky.social ! 🚀 🌐 sites.google.com/view/mmfm3rd... 📅 Deadline: Apr 1, 2025 (non-proceedings) 📝 Submission: cmt3.research.microsoft.com/MMFM2025
cmt3.research.microsoft.com
Conference Management Toolkit - Login
Microsoft's Conference Management Toolkit is a hosted academic conference management system. Modern interface, high scalability, extensive features and outstanding support are the signatures of Micros...
023
Reposted by @julianpollmann.bsky.social
Dr. Nicole Göbel @creativity.bsky.social · 26/02/2025
⏳️ Advice from a grandchild of the Nazi generation to the US: Don’t wait until it’s too late—organize now: generalstrikeus.com
generalstrikeus.com
The General Strike
The General strike -the people united shall never be stopped. If we strike together we can make true real change. Sign the strike card today.
45719
Reposted by @julianpollmann.bsky.social
Jörg Spengler 🇪🇺 @joergspengler.bsky.social · 15/03/2025
Wir reden zu viel über Schulden und zu wenig über Verteilungsgerechtigkeit. Unsere Gesellschaft krankt an der riesigen Schere zwischen arm und reich. #TaxTheRich #Vermögenssteuer
Graphik: Das reichste 1 Prozent der Weltbevölkerung besitzt 43,3 Prozent des gesamten Vermögens. Quelle: Credit Suisse 2020
281016331
julianpollmann.bsky.social @julianpollmann.bsky.social · 13/03/2025
@me-datapoint.bsky.social on molecular Fingerprints during the #matchms workshop @zdd-hsd.bsky.social
021
Reposted by @julianpollmann.bsky.social
Florian Huber @me-datapoint.bsky.social · 12/03/2025
#matchms workshop in full swing! Great fun to work with this fantastic group of people on improving, expanding, applying matchms for handling #massspec data in #Python.
1102
Reposted by @julianpollmann.bsky.social
Bernhard Suchy @toscaput.bsky.social · 01/03/2025
Wie wir eigentlich das Karnevals-Wochenende verbringen sollten.
282
julianpollmann.bsky.social @julianpollmann.bsky.social · 27/02/2025
One common issue i noticed from chatting to other #RSEs: short-term contracts prevent success of software projects and good research (even though I'm more privileged with a longer contract). #deRSE25 #IchBinHanna
141
julianpollmann.bsky.social @julianpollmann.bsky.social · 27/02/2025
Really enjoyed my first conference on research software #deRSE25 @kit.edu. Thanks for all the nice insights, talks and the good organization.
000
julianpollmann.bsky.social @julianpollmann.bsky.social · 12/02/2025
Oh nice, #HuggingFace released a new AI Agents Course: huggingface.co/learn/agents...
huggingface.co
Welcome to the 🤗 AI Agents Course - Hugging Face Agents Course
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
000
Reposted by @julianpollmann.bsky.social
Ethan Mollick @emollick.bsky.social · 02/02/2025
Been waiting for someone to test this and see if it works - can multiple AI agents fact-checking each other reduce hallucinations? The answer appears to be yes - using 3 agents with a structured review process reduced hallucination scores by 96% across 310 test cases. arxiv.org/pdf/2501.13946
813725
Reposted by @julianpollmann.bsky.social
Sebastian Friedrich @sebfriedrich.bsky.social · 01/02/2025
Ständig ist von der „Mehrheit der Bevölkerung“ die Rede, die dieses oder jenes will. Nur seltsam, dass der Wunsch nach Umverteilung dabei fast nie erwähnt wird.
Politbarometer nach dem 70 Prozent eine stärkere Besteuerung hoher Einkommen wollen
732453876
Reposted by @julianpollmann.bsky.social
Anna Berg @bergdame.bsky.social · 30/01/2025
Sie hat es tatsächlich getan. Was nur mehr zeigt, wie unglaublich ist, das gestern passierte. #merkel www.buero-bundeskanzlerin-ad.de/erklaerungen...
buero-bundeskanzlerin-ad.de
Erklärung von Bundeskanzlerin a. D. Dr. Angela Merkel zur Abstimmung im Deutschen Bundestag am 29. Januar 2025
118044
Reposted by @julianpollmann.bsky.social
tante @tante.cc · 27/01/2025
With all the hype around DeepSeek-r1 it's important to remark that is is still an LLM with all those properties (trained on data taken without consent/license, full of fabrications).
3497
Reposted by @julianpollmann.bsky.social
Lewis Tunstall @lewtun.bsky.social · 25/01/2025
We are reproducing the full DeepSeek R1 data and training pipeline so everybody can use their recipe. Instead of doing it in secret we can do it together in the open! Follow along: github.com/huggingface/...
github.com
GitHub - huggingface/open-r1: Fully open reproduction of DeepSeek-R1
Fully open reproduction of DeepSeek-R1. Contribute to huggingface/open-r1 development by creating an account on GitHub.
620036
julianpollmann.bsky.social @julianpollmann.bsky.social · 28/01/2025
Nice article on how #DeepSeek R1 was trained: www.philschmid.de/deepseek-r1
philschmid.de
Bite: How Deepseek R1 was trained
5 Minute Read on how Deepseek R1 was trained using Group Relative Policy Optimization (GRPO) and RL-focused multi-stage training approach.
000
Reposted by @julianpollmann.bsky.social
Amrei Bahr @amreibahr.bsky.social · 28/01/2025
Ich meine das mit dem Aufruf zum Wählen übrigens ernst. Laut Forsa könnten wir 28% (!) Nicht-Wähler_innen haben. Liebe Wissenschaftler_innen, liebe Wissenschaftsinstitutionen: Euch hören viele zu. Erinnert sie, wie wichtig Wählen ist. Für unsere Demokratie. & ermuntert sie, Botschaft weiterzutragen!
14295116
Reposted by @julianpollmann.bsky.social
Zentrum für Politische Schönheit @politicalbeauty.bsky.social · 24/01/2025
Kauft keine Teslas! Kooperation mit @ledbydonkeys.org
17825162
Reposted by @julianpollmann.bsky.social
netzpolitik.org @netzpolitik.org · 24/01/2025
Bald will der IT-Dienstleister #Doctolib die Daten seiner Nutzer:innen zum Training sogenannter Künstlicher Intelligenz einsetzen. Wenn sie einwilligen, sollen auch ihre Gesundheitsdaten dafür genutzt werden. Das geht aus aktualisierten Datenschutzhinweisen hervor. netzpolitik.org/2025/neue-da...
netzpolitik.org
Neue Datenschutzhinweise: Doctolib will KI-Modelle mit Gesundheitsdaten trainieren
Ab Ende Februar will der IT-Dienstleister Doctolib die Daten seiner Nutzer:innen für das Training sogenannter Künstlicher Intelligenz einsetzen. Wenn sie einwilligen, sollen auch ihre Gesundheitsdaten...
89960
Reposted by @julianpollmann.bsky.social
Carolyn Anderson @carolynanderson.bsky.social · 23/01/2025
excited to say that our Substance Beats Style paper was accepted to NAACL! We investigate *why* student-written programming prompts don't work well for LLMs, and find that while students think it's because of technical vocabulary gaps, it's actually information content that matters
arxiv.org
Substance Beats Style: Why Beginning Students Fail to Code with LLMs
Although LLMs are increasing the productivity of professional programmers, existing work shows that beginners struggle to prompt LLMs to solve text-to-code tasks. Why is this the case? This paper expl...
3348
Reposted by @julianpollmann.bsky.social
Inge Rosenberger @ingerosenberger.bsky.social · 21/01/2025
Politik sollte man den Profis überlassen. (frei nach CL)
tagesspiegel.de
Ungültige Aufstellung in 14 Landesverbänden?: Landeswahlleiter prüfen FDP-Listen zur Bundestagswahl
Die FDP kämpft um den Wiedereinzug in den Bundestag. Nun legt ein Parteimitglied Beschwerde gegen die Listenaufstellung in vierzehn Landesverbänden ein. In NRW reagierte die Partei bereits.
1022647