Fredrik Johansson
@frejohk.bsky.social
Associate professor, Chalmers University of Technology. Machine learning for decision making & healthcare. http://healthyai.se, http://fredjo.com
Key to this is to decouple environment interaction from language generation while maintaining the reasoning capabilities of pre-trained models.
Project page: expa-rl.github.io
Pre-print: arxiv.org/abs/2510.07581
PS. Nick is on the job market!
Project page: expa-rl.github.io
Pre-print: arxiv.org/abs/2510.07581
PS. Nick is on the job market!
EARL
Expanding the Action Space of LLMs to Reason Beyond Language
expa-rl.github.io
October 21, 2025 at 8:34 AM
Key to this is to decouple environment interaction from language generation while maintaining the reasoning capabilities of pre-trained models.
Project page: expa-rl.github.io
Pre-print: arxiv.org/abs/2510.07581
PS. Nick is on the job market!
Project page: expa-rl.github.io
Pre-print: arxiv.org/abs/2510.07581
PS. Nick is on the job market!
Thank you Branislav Kveton, Sandeep Juneja, Slawomir Nowaczyk and Yevgeny Seldin for serving as Newton's grading committee and opponent!
Read Newton's PhD thesis here: research.chalmers.se/publication/...
Read Newton's PhD thesis here: research.chalmers.se/publication/...
Leveraging Structural Priors and Historical Data for Practical Treatment Personalization with Multi-Armed Bandits
Personalizing treatments for patients often requires sequentially trying different options from a set of available therapies until the most effective one is identified for the patient’s characteristic...
research.chalmers.se
October 13, 2025 at 1:48 PM
Thank you Branislav Kveton, Sandeep Juneja, Slawomir Nowaczyk and Yevgeny Seldin for serving as Newton's grading committee and opponent!
Read Newton's PhD thesis here: research.chalmers.se/publication/...
Read Newton's PhD thesis here: research.chalmers.se/publication/...
Anton's thesis: research.chalmers.se/publication/...
Lena's thesis: research.chalmers.se/en/publicati...
Lena's thesis: research.chalmers.se/en/publicati...
Interpretable Machine Learning for Modeling, Evaluating, and Refining Clinical Decision-Making
Machine learning offers great promise for developing new treatment policies from observational clinical data. However, a key challenge in this offline setting is reliably assessing the performance of ...
research.chalmers.se
August 29, 2025 at 12:18 PM
Anton's thesis: research.chalmers.se/publication/...
Lena's thesis: research.chalmers.se/en/publicati...
Lena's thesis: research.chalmers.se/en/publicati...
In a tree, features are used depending on the values and availability of other features. By penalizing reliance on missing values in computing predictions, we can learn trees that are unlikely to ask for a feature unless it is available (and predictive)
May 8, 2025 at 6:36 AM
In a tree, features are used depending on the values and availability of other features. By penalizing reliance on missing values in computing predictions, we can learn trees that are unlikely to ask for a feature unless it is available (and predictive)
In this work, we show that you can fit interpretable models that are unlikely to even *ask for* features that are missing at test time by learning to avoid using them in the first place. For trees, this is contextual...
May 8, 2025 at 6:36 AM
In this work, we show that you can fit interpretable models that are unlikely to even *ask for* features that are missing at test time by learning to avoid using them in the first place. For trees, this is contextual...
If you want to fit an interpretable ML model like a tree or a GLM, how can you make sure that it's still accurate and interpretable when input features are missing at test time? Standard methods like imputation or missingness indicators tend to break one or the other...
May 8, 2025 at 6:36 AM
If you want to fit an interpretable ML model like a tree or a GLM, how can you make sure that it's still accurate and interpretable when input features are missing at test time? Standard methods like imputation or missingness indicators tend to break one or the other...
Our CS department is listing 5 positions. Make sure to select "F. Johansson - Reducing waste in tabular machine learning by generalizing out-of-table" to indicate which project you would like to work in.
Deadline May 15!
Deadline May 15!
April 23, 2025 at 6:16 AM
Our CS department is listing 5 positions. Make sure to select "F. Johansson - Reducing waste in tabular machine learning by generalizing out-of-table" to indicate which project you would like to work in.
Deadline May 15!
Deadline May 15!
Our CS department is listing 5 positions. Make sure to select "F. Johansson - Reducing waste in tabular machine learning by generalizing out-of-table" to indicate which project you would like to work in.
Deadline May 15!
Deadline May 15!
April 23, 2025 at 6:15 AM
Our CS department is listing 5 positions. Make sure to select "F. Johansson - Reducing waste in tabular machine learning by generalizing out-of-table" to indicate which project you would like to work in.
Deadline May 15!
Deadline May 15!
No argument there—I'm pretty sure that's a different question though.
March 4, 2025 at 5:10 PM
No argument there—I'm pretty sure that's a different question though.
Clearly not but I was surprised at how many orals at NeurIPS were missing speakers too. Any chance of having reserve speakers? Sure, there is an honor in being awarded an oral but, as a conference-goer, having a 20-min gap in the schedule due to a missing speaker seems like a missed opportunity.
March 4, 2025 at 3:26 PM
Clearly not but I was surprised at how many orals at NeurIPS were missing speakers too. Any chance of having reserve speakers? Sure, there is an honor in being awarded an oral but, as a conference-goer, having a 20-min gap in the schedule due to a missing speaker seems like a missed opportunity.
The first paper by Herman Bergström, Emil Carlsson, Devdatt Dubhashi, and me, explores how important context is to active preference learning from pairwise feedback: arxiv.org/abs/2405.03059
In the second, I develop a benchmark for evaluating estimators of causal effects. arxiv.org/abs/2405.16069
In the second, I develop a benchmark for evaluating estimators of causal effects. arxiv.org/abs/2405.16069
IncomeSCM: From tabular data set to time-series simulator and causal estimation benchmark
Evaluating observational estimators of causal effects demands information that is rarely available: unconfounded interventions and outcomes from the population of interest, created either by randomiza...
arxiv.org
December 10, 2024 at 4:25 PM
The first paper by Herman Bergström, Emil Carlsson, Devdatt Dubhashi, and me, explores how important context is to active preference learning from pairwise feedback: arxiv.org/abs/2405.03059
In the second, I develop a benchmark for evaluating estimators of causal effects. arxiv.org/abs/2405.16069
In the second, I develop a benchmark for evaluating estimators of causal effects. arxiv.org/abs/2405.16069