Skip to content

Commit

Permalink
Update
Browse files Browse the repository at this point in the history
  • Loading branch information
@guokan-shang
guokan-shang committed Jan 24, 2025
1 parent ddb6d9d commit a310713
Show file tree
Hide file tree
Showing 5 changed files with 125 additions and 108 deletions.
78 changes: 41 additions & 37 deletions program-day-1/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -70,23 +70,23 @@ <h2>Program on Wednesday, February 12</h2>
</tr>
</table>

<table id="PUTSPEAKERNAMEHERE">
<table class="plenary">
<tr>
<td class="date" rowspan="3">
9:30am
09:30am
</td>
<td class="title">
Factuality Challenges in the Era of Large Language Models: Can we Keep LLMs Safe and Factual?
Opening Remarks
</td>
</tr>
<tr>
<td class="speaker">
<a target="_blank" href="https://mbzuai.ac.ae/study/faculty/preslav-nakov/">Preslav Nakov</a> (MBZUAI)
<a target="_blank" href="https://mbzuai.ac.ae/study/faculty/professor-eric-xing/">Eric Xing</a> (MBZUAI & Carnegie Mellon University)
</td>
</tr>
<tr>
<td class="abstract">
We will discuss the risks, the challenges, and the opportunities that Large Language Models (LLMs) bring regarding factuality. We will then delve into our recent work on using LLMs for fact-checking, on detecting machine-generated text, and on fighting the ongoing misinformation pollution with LLMs. We will also discuss work on safeguarding LLMs, and the safety mechanisms we incorporated in Jais-chat, the world's best open Arabic-centric foundation and instruction-tuned LLM, based on our Do-Not-Answer dataset. Finally, we will present a number of LLM fact-checking tools recently developed at MBZUAI: (i) LM-Polygraph, a tool to predict an LLM's uncertainty in its output using cheap and fast uncertainty quantification techniques, (ii) Factcheck-Bench, a fine-grained evaluation benchmark and framework for fact-checking the output of LLMs, (iii) Loki, an open-source tool for fact-checking the output of LLMs, developed based on Factcheck-Bench and optimized for speed and quality, (iv) OpenFactCheck, a framework for fact-checking LLM output, for building customized fact-checking systems, and for benchmarking LLMs for factuality, and (v) LLM-DetectAIve, a tool for machine-generated text detection.
<!-- Kept for Spacing -->
</td>
</tr>
</table>
Expand All @@ -97,17 +97,17 @@ <h2>Program on Wednesday, February 12</h2>
10:10am
</td>
<td class="title">
TBD
Statistical Methods for Assessing the Factual Accuracy of Large Language Models
</td>
</tr>
<tr>
<td class="speaker">
<a target="_blank" href="#">TBD</a> (TBD)
<a target="_blank" href="https://candes.su.domains/">Emmanuel Candès</a> (Stanford University)
</td>
</tr>
<tr>
<td class="abstract">
TBD
We present new statistical methods for obtaining validity guarantees on the output of large language models (LLMs). These methods enhance conformal prediction techniques to filter out claims/remove hallucinations while providing a finite-sample guarantee on the error rate of what it being presented to the user. This error rate is adaptive in the sense that it depends on the prompt to preserve the utility of the output by not removing too many claims. We demonstrate performance on real-world examples. This is joint work with John Cherian and Isaac Gibbs.
</td>
</tr>
</table>
Expand Down Expand Up @@ -139,7 +139,7 @@ <h2>Program on Wednesday, February 12</h2>
</tr>
<tr>
<td class="speaker">
<a target="_blank" href="https://velcin.github.io/">Julien Velcin</a> (Université de Lyon)
<a target="_blank" href="https://velcin.github.io/">Julien Velcin</a> (University of Lyon)
</td>
</tr>
<tr>
Expand All @@ -155,42 +155,41 @@ <h2>Program on Wednesday, February 12</h2>
11:40am
</td>
<td class="title">
Exploiting Knowledge for Model-based Deep Music Generation
Foundation Models in Biology, from Histo-pathology to Genomics
</td>
</tr>
<tr>
<td class="speaker">
<a target="_blank" href="https://www.telecom-paris.fr/gael-richard">Gaël Richard</a> (Télécom Paris)
<a target="_blank" href="https://jpvert.github.io/">Jean-Philippe Vert</a> (Owkin)
</td>
</tr>
<tr>
<td class="abstract">
We will describe and illustrate the concept of hybrid (or model-based) deep learning for music generation. This paradigm refers here to models that associates data-driven and model-based approaches in a joint framework by integrating our prior knowledge about the data in more controllable deep models. In the music domain, prior knowledge can relate for instance to the production or propagation of sound (using an acoustic or physical model) or how music is composed or structured (using a musicological model). In this presentation, we will first illustrate the concept and potential of such model-based deep learning approaches and then describe in more details its application to unsupervised music separation with source production models, music timbre transfer with diffusion and symbolic music generation with transformers using structured informed positional encoding.
Large self-supervised foundation models have boosted the capabilities of AI models in natural language processing and computer vision. Can they also boost our understanding of biology and help us improve diagnosis and find new treatments for diseases like cancer? In this talk I will present our efforts to train foundation models for histopathology images of tissues, and to connect visual observations to the underlying genomics of the cells, paving the way to biomedical innovation in diagnosis and precision medicine.
</td>
</tr>
</table>

<!--<table class="plenary">
<table id="PUTSPEAKERNAMEHERE">
<tr>
<td class="date" rowspan="3">
11:00am
12:20am
</td>
<td class="title">
This is a Plenary Talk!
Exploiting Knowledge for Model-based Deep Music Generation
</td>
</tr>
<tr>
<td class="speaker">
Plenary Speaker
<a target="_blank" href="https://www.telecom-paris.fr/gael-richard">Gaël Richard</a> (Télécom Paris)
</td>
</tr>
<tr>
<td class="abstract">
Here&rsquo;s the abstract for the plenary talk! Notice again that the formatting of this time-block is a bit different that the rest of of the talks.
Filler text: Lorem ipsum dolor sit amet, consectetur adipiscing elit. Vestibulum dignissim et est et euismod. Fusce et metus tempus, pellentesque ex at, convallis nulla. Ut fringilla commodo tincidunt. Fusce sed est eu massa placerat iaculis eu at mauris. Nullam ut mollis nisi, quis malesuada risus. Interdum et malesuada fames ac ante ipsum primis in faucibus. Nam ipsum tortor, suscipit non tincidunt vel, bibendum in libero. Nulla facilisi. Pellentesque vitae neque metus. Cras quis est pharetra, vestibulum nisl et, viverra ipsum. Etiam porta dignissim purus, quis tempor metus volutpat eu. Praesent pulvinar libero eget purus tincidunt finibus.
We will describe and illustrate the concept of hybrid (or model-based) deep learning for music generation. This paradigm refers here to models that associates data-driven and model-based approaches in a joint framework by integrating our prior knowledge about the data in more controllable deep models. In the music domain, prior knowledge can relate for instance to the production or propagation of sound (using an acoustic or physical model) or how music is composed or structured (using a musicological model). In this presentation, we will first illustrate the concept and potential of such model-based deep learning approaches and then describe in more details its application to unsupervised music separation with source production models, music timbre transfer with diffusion and symbolic music generation with transformers using structured informed positional encoding.
</td>
</tr>
</table>-->
</table>

<table>
<tr>
Expand All @@ -208,18 +207,20 @@ <h2>Program on Wednesday, February 12</h2>
</tr>
</table>



<table id="PUTSPEAKERNAMEHERE">
<tr>
<td class="date" rowspan="3">
13:30pm
14:00am
</td>
<td class="title">
TBD
TBD
</td>
</tr>
<tr>
<td class="speaker">
<a target="_blank" href="https://www.linkedin.com/in/thomas-pierrot-120a43128/">Thomas Pierrot</a> (InstaDeep)
<a target="_blank" href="https://keg.cs.tsinghua.edu.cn/jietang/">Jie Tang</a> (Tsinghua University)
</td>
</tr>
<tr>
Expand All @@ -229,27 +230,30 @@ <h2>Program on Wednesday, February 12</h2>
</tr>
</table>


<table id="PUTSPEAKERNAMEHERE">
<tr>
<td class="date" rowspan="3">
14:10pm
14:40pm
</td>
<td class="title">
Towards the Alignment of Geometric and Text Latent Spaces
TBD
</td>
</tr>
<tr>
<td class="speaker">
<a target="_blank" href="https://www.lix.polytechnique.fr/~maks/">Maks Ovsjanikov</a> (École Polytechnique)
<a target="_blank" href="https://misovalko.github.io/index.html">Michal Valko</a> (INRIA & Stealth Startup)
</td>
</tr>
<tr>
<td class="abstract">
Recent works have shown that, when trained at scale, uni-modal 2D vision and text encoders converge to learned features that share remarkable structural properties, despite arising from different representations. However, the role of 3D encoders with respect to other modalities remains unexplored. Furthermore, existing 3D foundation models that leverage large datasets are typically trained with explicit alignment objectives with respect to frozen encoders from other representations. In this talk I will discuss some results on the alignment of representations obtained from uni-modal 3D encoders compared to text-based feature spaces. Specifically, I will show that it is possible to extract subspaces of the learned feature spaces that have common structure between geometry and text. This alignment also leads to improvement in downstream tasks, such as zero shot retrieval. Overall, this work helps to highlight both the shared and unique properties of 3D data compared to other representations.
TBD
</td>
</tr>
</table>



<table>
<tr>
<td class="date" rowspan="2">
Expand All @@ -269,15 +273,15 @@ <h2>Program on Wednesday, February 12</h2>
<table id="PUTSPEAKERNAMEHERE">
<tr>
<td class="date" rowspan="3">
15:00pm
15:30pm
</td>
<td class="title">
Moshi: A Speech-text Foundation Model for Real-time Dialogue
</td>
</tr>
<tr>
<td class="speaker">
<a target="_blank" href="https://ai.honu.io/">Alexandre Desfossez</a> (Kyutai)
<a target="_blank" href="https://ai.honu.io/">Alexandre Défossez</a> (Kyutai)
</td>
</tr>
<tr>
Expand All @@ -290,41 +294,41 @@ <h2>Program on Wednesday, February 12</h2>
<table id="PUTSPEAKERNAMEHERE">
<tr>
<td class="date" rowspan="3">
15:40pm
16:10pm
</td>
<td class="title">
TBD
Feature-Conditioned Graph Generation using Latent Diffusion Models
</td>
</tr>
<tr>
<td class="speaker">
<a target="_blank" href="#">TBD</a> (TBD)
<a target="_blank" href="https://users.uop.gr/~nikolentzos/">Giannis Nikolentzos</a> (University of Peloponnese)
</td>
</tr>
<tr>
<td class="abstract">
TBD
Graph generation has emerged as a crucial task in machine learning, with significant challenges in generating graphs that accurately reflect specific properties. In this talk, I will present Neural Graph Generator, our recently released model which utilizes conditioned latent diffusion models for graph generation. The model employs a variational graph autoencoder for graph compression and a diffusion process in the latent vector space, guided by vectors summarizing graph statistics. Overall, this work represents a shift in graph generation methodologies, offering a more practical and efficient solution for generating diverse graphs with specific characteristics.
</td>
</tr>
</table>

<table class="plenary">
<table id="PUTSPEAKERNAMEHERE">
<tr>
<td class="date" rowspan="3">
16:20pm
16:50pm
</td>
<td class="title">
TBD
Redefining AI Reasoning: From Self-Guided Exploration to Causal Loops, and Transformer-GNN Fusion
</td>
</tr>
<tr>
<td class="speaker">
<a target="_blank" href="https://mbzuai.ac.ae/study/faculty/professor-eric-xing/">Eric Xing</a> (MBZUAI)
<a target="_blank" href="https://mbzuai.ac.ae/study/faculty/martin-takac/">Martin Takáč</a> (MBZUAI)
</td>
</tr>
<tr>
<td class="abstract">
TBD
In this talk, we explore three intertwined directions that collectively redefine how AI systems reason about complex tasks. First, we introduce Self-Guided Exploration (SGE), a prompting strategy that enables Large Language Models (LLMs) to autonomously generate multiple “thought trajectories” for solving combinatorial problems. Through iterative decomposition and refinement, SGE delivers significant performance gains on NP-hard tasks—showcasing LLMs’ untapped potential in reasoning, logistics and resource management problems. Next, we delve into the Self-Referencing Causal Cycle (ReCall), a mechanism that sheds new light on LLMs’ ability to recall prior context from future tokens. Contrary to the common belief that unidirectional token generation fundamentally restricts memory, ReCall illustrates how “cycle tokens” create loops in the training data, enabling models to overcome the notorious “reversal curse.” Finally, we present a Transformer-GNN fusion architecture that addresses Transformers’ limitations in processing graph-structured data.
</td>
</tr>
</table>
Expand Down
Loading

0 comments on commit a310713

Please sign in to comment.