Goodbye Vancouver! 👋

What an unforgettable experience at #NeurIPS2024 in beautiful Vancouver! 🎉 Grateful for the chance to present my paper, attend amazing talks, and connect with brilliant minds. The ideas, energy, and city vibes made it truly special. Until next time, @NeurIPSConf !

🔗 Paper: .🔗 Code: 🔗 Poster: #XAI #ExplainableAI #explainability #Interpretability.

Join us tomorrow at our poster! 🌟. I’ll present our work, "MambaLRP: Explaining Selective State Space Sequence Models," at #NeurIPS2024. 📅 Thursday, December 12 🕟 4:30 PM PST 📍 East Exhibit Hall A-C, Poster #2804. Looking forward to seeing you there!

Excited to be at #NeurIPS2024 this week! I'll present our work, "MambaLRP: Explaining Selective State Space Sequence Models," on Thu 12 Dec 4:30 PM PST (Poster #2804). Let’s connect to chat about XAI & Interpretability. See you there!.

Curious to learn more? Check out our paper and GitHub repo here:.[Paper]: .[Code]: #XAI #explainability #LLM @XAI_Research #NeurIPS #interpretability.

2. Uncovering undesired decision strategies in image classification. 3. Identifying gender bias in language models. 4. Analyzing the long-range capabilities of Mamba. 5. Introducing an explanation-aware retrieval accuracy metric for the needle-in-a-haystack test.

#FUBerlin, #KoreaUniversity , and #MPI Saarbrücken. Our extensive evaluations demonstrate MambaLRP’s superior performance in explaining model predictions compared to existing methods. Key highlights include:.1. Gaining insights into the model’s prediction mechanisms.

🎉 Thrilled to announce that our paper MambaLRP has been accepted at #NeurIPS2024! .Explain your Mamba models in a faithful, efficient, and effective way! 🐍.Huge thanks to @EberleOliver, Grégoire Montavon, and Klaus-Robert Müller @bifoldberlin, @ml_tuberlin , @GoogleDeepMind,

Absolutely! It’s heartwarming to see that the effort put into addressing the reviewers’ comments has been recognized and has opened the door for a friendly and open discussion. Regardless of the final decision, it’s great to know that the hard work has not been in vain.

Very excited to be in beautiful Vienna 🇦🇹 for #ICML2024. I look forward to meeting and connecting with others working on interpretability and XAI topics.

At last, I would like to express my gratitude to all my collaborators: @EberleOliver, Grégoire Montavon, and Klaus-Robert Müller. @bifoldberlin @ml_tuberlin @TUBerlin @GoogleDeepMind.#Explainable_AI #Mamba #Interpretability. 7/7.

We further show the usefulness of MambaLRP in several areas: gaining concrete insights into the model's prediction mechanism, uncovering undesired decision strategies and biases, and analyzing the long-range capabilities of Mamba. 6/7.

Our proposed method is theoretically sound and excels in achieving state-of-the-art explanation performance across a diverse range of models and datasets. 5/7

Guided by the axiom of relevance conservation, we identify specific components in the Mamba architecture, which cause unfaithful explanations. To remedy this issue, we propose MambaLRP, which ensures a more stable and reliable relevance propagation through these components. 4/7.

To foster the reliable use of Mamba models in real-world scenarios, it is crucial to augment their transparency. Our work bridges this critical gap by bringing explainability, particularly Layer-wise Relevance Propagation (LRP), to the Mamba architecture. 3/7.

Paper: Code: 2/7

📢 New Preprint Alert 🚨.Ever Wondered How to Make Mamba Models More Transparent? . Meet MambaLRP🔥.We introduce MambaLRP, a novel algorithm within the Layer-wise Relevance Propagation (LRP) framework, specially designed to enhance the explainability of Mamba models. 1/7

RT @MSFTResearch: The 4th International Workshop on Large Scale Holistic Video Understanding in Conjunction with @CVPR is accepting paper s….

RT @bifoldberlin: Full house: >100 international scientists joined us today for the #BIFOLD Research-Symposium @FDT_Berlin. Scientific keyn….