RT @DMLRJournal: 'Potion: Towards Poison Unlearning' by Stefan Schoepf, Jack Foster, Alexandra Brintrup Action Editor: Hongyang Zhang ht…

🛜 Turns out you can get GPT o1-mini to use the web browser by directly invoking it via the hidden invocation actions that the model can call itself. When asked without direct invocation you will just get the knowledge cutoff message but with it you get full access.

I got tired of searching for authorship positions on Scholar and made an automatic highlighting Chrome extension: Hope it helps you too to quickly focus on the main papers an author contributed to

RT @ShashwatGoel7: You find some training data sources were compromised, potentially causing #backdoors, #bias, #mislabels etc. Can you rem…

@ShashwatGoel7 @AmyPrb @AmartyaSanyal @ponguru @OxfordTVG Awesome work! Happy to see our SSD method in action We also have a parameter tuning free version to save you some tuning work in the future 😉:

#AAAI24 poster session ✅ 📄 Fast Machine Unlearning Without Retraining Through Selective Synaptic Dampening 🔗 You can find the code for our #machine #unlearning paper here:

RT @jaschasd: Have you ever done a dense grid search over neural network hyperparameters? Like a *really dense* grid search? It looks like…

Machine unlearning can help to fix models that were trained on data with errors. But finding the right parameters for new unlearning tasks is hard. -> Parameter-Tuning-Free Data Entry Error Unlearning with Adaptive Selective Synaptic Dampening

@jet2tweets How come every other airline can fly out of Innsbruck today except you? ❄️

Our #machine #unlearning paper got accepted at #AAAI 2024! If you know of any cool companies looking for unlearning interns/collabs, let me know! Fast Machine Unlearning Without Retraining Through Selective Synaptic Dampening: @RealAAAI @JackFosterML

RT @JackFosterML: How can we FORGET private information from large #MachineLearning models? Excited to share our newest preprint addressin…

How can we remove private information from large machine-learning models? Our answer: Fast #MachineUnlearning Without Retraining Through Selective Synaptic Dampening ArXiv:

@solomonwalker Good timing, just uploaded our preprint: Fast Machine Unlearning Without Retraining Through Selective Synaptic Dampening Arxiv:

Happy to share the preprint of my #RL for #combinatorial #optimization paper. I’ll also present my research on this topic at the #IJCAI WoSePCO Workshop in a few weeks. Combining vehicle routing and constrained 3D-bin packing: