Bar plot comparing fuzzing strategies. y-axis: combined violation and diversity score. Each bar represents an average of normalized violations and normalized diversity score, equally weighted. Our metrics are shown in green, with our two best VO-KMVP and VO-KMOC highlighted in dark green. Neuron coverage metrics are shown in purple, and basic metrics in red. Conventional testing (model test set with no mutations) is shown in blue. Our methods reach scores above 0.8, whereas conventional testing sits at 0.1

Our approach borrows an idea from software verification: coverage-guided fuzzing.

We systematically mutate inputs and search for stimulation patterns that violate biophysical constraints - uncovering diverse safety violations that conventional testing misses.

#Neurotech #MLResearch #AIVerification

We were thrilled to host @mtutek.bsky.social at our lab last week.
His talk "From Internals to Integrity: How Insights into Transformer LMs Improve Safety, Interpretability, and Explanation Faithfulness" led to great discussions! 👏
#Transformers #AISafety #ExplainableAI #MLResearch #NLProc

[AI] Propose, Solve, Verify: Self-Play Through Formal Verification
A Wilf, P Aggarwal, B Parno, D Fried... [CMU] (2025)
arxiv.org/abs/2512.18160
#AI
#MachineLearning
#DeepLearning
#ArtificialIntelligence
#DataScience
#NeuralNetworks
#MLResearch
#AIResearch

Learning Dynamics of LLM Finetuning Learning dynamics, which describes how the learning of specific training examples influences the model's predictions on other examples, gives us a powerful tool for understanding the behavior of deep…

New on arXiv: “Learning Dynamics of LLM Finetuning.” A unified view of SFT & DPO reveals a squeezing effect driving confidence decay in off-policy DPO—and a simple SFT tweak that boosts downstream wins. arxiv.org/abs/2407.10490 #LLM #RLHF #MLResearch @arxiv

Paper page - Adaptation of Agentic AI Join the discussion on this paper page

A clean framework for adapting agentic AI: adapt the agent or the tools, with signals from execution or outputs—yielding four practical paradigms + design guidance. Read the survey: huggingface.co/papers/2512.... #AIagents #LLM #MLresearch

#DigitalBehavior #MLResearch #NLPResearch #SocialMediaData

#CallForPapers: Our full-day workshop at The Web Conference 2026 #TheWebConf2026 invites submissions on reproducible and reusable computational approaches for social and #WebData.
👉 easychair.org/cfp/r2...

Deadline: Dec 18th, 2025

Two weeks left! Submit a 1-page overview for our AAAI “Perspectives from the Community” session alongside the workshop on the Rashomon Effect (Jan 27, 2026) ✨ Selected contributors will give 5–15 min talks.

Deadline Nov 30
➡️ forms.gle/h4Dj2MfVqYpa...

#AAAI26 #MLResearch #Uncertainty

While primarily a theoretical tool, discussions suggest central flow insights could lead to practical applications. This includes potentially speeding up convergence and improving future optimization methods. #MLResearch 4/6

Feature normalization in Ijepa helps stabilize training and improve performance in self-supervised learning. This project explores its impact on real-world vision tasks. Discover the insights: github.com/theAdamColton/elucidatin... #MLResearch #OpenSource

Excited to welcome Aditya Grover, UCLA, as a keynote speaker at #IJCAI2025 in #Montreal. He is the winner of the IJCAI-25 Computers and Thought Award: aditya-grover.github.io

#AI #GenerativeAI #MLResearch

How do researchers turn papers into real-world impact?

At the WiML Symposium @ ICML 2025, join Microsoft’s Ida Momennejad and Yuhang He for a roundtable on academic-industry collaboration.

🗓 July 16 | 1–2 PM PT | Vancouver

#WiML #ICML2025 #MLResearch

In this episode, we talk about:
🔹 Decentralized & collaborative AI
🔹 Scaling laws & fine-tuning tricks
🔹 Data attribution for LLMs
🔹 Multilingual learning
🔹 UK 🇬🇧 vs. Canada 🇨🇦 research ecosystems
/3

#MLResearch #LLM #DecentralizedAI

Until then, let’s call this what it is: intellectual laziness masquerading as empirical rigor.

📄 Read the paper if you must

#TimeSeries #MachineLearning #BadScience #SyntheticData #AISnakeOil #GPDelusions #MLResearch

✅ Plug-and-play with DDIM, Flow Matching, etc.

A minimalist yet powerful update to how we think about generative diffusion models.

📄 Link to paper -> openaccess.thecvf.co...

#CVPR2025 #DiffusionModels #KAN #GenerativeAI #SchrodingerBridge #MLResearch

[9/9]

Appreciate any advice, pointers to relevant papers, or even “don’t do this” cautionary tales.
Thanks in advance!

#transformers #sparsity #maskedmodeling #deeplearning #symbolicAI #mlresearch #attentionmodels #structureddata

#ConformalPrediction #SHAP #ExplainableAI #TrustworthyAI #XAI #MLBias #MLResearch #FeatureAttribution

#MachineLearning #DataPreparation #TargetTransformation #MLResearch #DataScience

🎉Hats off to Dr. Stefan Blücher! He defended his PhD yesterday🎓

Thesis: "Towards Scalable and Transparent ML Algorithms with Applications in Explainable AI and Quantum Chemistry"

Thanks to all supporters!
#XAI #QuantumML #MLResearch #PhDCompleted @tuberlin.bsky.social @fraunhoferhhi.bsky.social

NoProp: Training Neural Networks without Back-propagation or Forward-propagation The canonical deep learning approach for learning requires computing a gradient term at each layer by back-propagating the error signal from the output towards each learnable parameter. Given the…

NoProp: Training Neural Nets Without Backprop or Forwardprop
A new paper proposes "NoProp"—a radical approach to training neural networks without traditional back/forward propagation.
arxiv.org/abs/2503.24322
#AI #NeuralNetworks #MLResearch

How do you handle imbalance in your ML workflows?

Drop your thoughts in the comments! Let's start a conversation about responsible and effective machine learning practices. 💬👇

onlinelibrary.wiley....

#MachineLearning #DataScience #AI #SMOTE #ImbalancedData #MLResearch #PredictiveModeling

Let’s demand real comparisons and rigorous benchmarks.

Otherwise, we’re just building castles in the air.

#TimeSeries #Transformers #NBEATS #MLResearch #DataScience #IntegrityMatters

Language Models are Few-Shot Learners, page 33 mentions bidirectional models as a promising direction. Any recent movement in that direction? #ai #mlresearch

if it doesn't have pretty pictures I don't want it.

#mlresearch #academicsky #machinelearning #research

Trying to wrap up all your random papers in one thesis somehow 🙃

#academicsky #phdlife #mlresearch

I’m looking forward to engaging discussions, insightful questions, and connecting with fellow researchers. If you’re attending NeurIPS, stop by my sessions—I’d love to chat! 🚀

#NeurIPS2024 #MachineLearning #ComputerVision #ProbabilisticModels #ErrorRecognition #AIResearch #MLResearch

This is #MLresearch. (Not even being sarcastic too.)

I see so many new people joining BlueSky! Super excited to see this place growing. Looking forward to a stable #ml and #mlresearch community here 🤖