Skip to content

Commit

Permalink
updated preprint link
Browse files Browse the repository at this point in the history
  • Loading branch information
@tobiasgerstenberg
tobiasgerstenberg committed Oct 8, 2024
1 parent 47449fa commit bd32ed5
Show file tree
Hide file tree
Showing 19 changed files with 236 additions and 82 deletions.
2 changes: 1 addition & 1 deletion content/publication/du2024robotic.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -17,7 +17,7 @@ image_preview = ""
selected = false
projects = []
#url_pdf = "papers/du2024robotic.pdf"
url_preprint = "https://arxiv.org/pdf/2406.15917"
url_preprint = "https://arxiv.org/abs/2406.15917"
url_code = ""
url_dataset = ""
url_slides = ""
Expand Down
8 changes: 4 additions & 4 deletions docs/404.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -237,6 +237,10 @@ <h1>Page not found</h1>

<h2>Publications</h2>

<ul>
<li><a href="https://cicl.stanford.edu/publication/franken2024sami/">Self-supervised alignment with mutual information: Learning to follow principles without preference labels</a></li>
</ul>

<ul>
<li><a href="https://cicl.stanford.edu/publication/jin2024marple/">MARPLE: A Benchmark for Long-Horizon Inference</a></li>
</ul>
Expand All @@ -253,10 +257,6 @@ <h2>Publications</h2>
<li><a href="https://cicl.stanford.edu/publication/du2024robotic/">To Err is Robotic: Rapid Value-Based Trial-and-Error during Deployment</a></li>
</ul>

<ul>
<li><a href="https://cicl.stanford.edu/publication/gerstenberg2024counterfactual/">Counterfactual simulation in causal cognition</a></li>
</ul>




Expand Down
2 changes: 1 addition & 1 deletion docs/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -110,7 +110,7 @@
<meta property="og:description" content="">
<meta property="og:locale" content="en-us">

<meta property="og:updated_time" content="2024-10-04T00:00:00&#43;00:00">
<meta property="og:updated_time" content="2024-10-08T00:00:00&#43;00:00">



Expand Down
20 changes: 10 additions & 10 deletions docs/index.xml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,9 +6,18 @@
<generator>Hugo -- gohugo.io</generator>
<language>en-us</language>
<copyright>&amp;copy; 2024 Tobias Gerstenberg</copyright>
<lastBuildDate>Fri, 04 Oct 2024 00:00:00 +0000</lastBuildDate>
<lastBuildDate>Tue, 08 Oct 2024 00:00:00 +0000</lastBuildDate>
<atom:link href="/" rel="self" type="application/rss+xml" />

<item>
<title>Self-supervised alignment with mutual information: Learning to follow principles without preference labels</title>
<link>https://cicl.stanford.edu/publication/franken2024sami/</link>
<pubDate>Tue, 08 Oct 2024 00:00:00 +0000</pubDate>

<guid>https://cicl.stanford.edu/publication/franken2024sami/</guid>
<description></description>
</item>

<item>
<title>MARPLE: A Benchmark for Long-Horizon Inference</title>
<link>https://cicl.stanford.edu/publication/jin2024marple/</link>
Expand Down Expand Up @@ -135,14 +144,5 @@
<description></description>
</item>

<item>
<title>STaR-GATE: Teaching Language Models to Ask Clarifying Questions</title>
<link>https://cicl.stanford.edu/publication/andukuri2024stargate/</link>
<pubDate>Sun, 31 Mar 2024 00:00:00 +0000</pubDate>

<guid>https://cicl.stanford.edu/publication/andukuri2024stargate/</guid>
<description></description>
</item>

</channel>
</rss>
49 changes: 48 additions & 1 deletion docs/member/tobias_gerstenberg/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -356,6 +356,53 @@ <h2 id="publications">Publications</h2>


<div class="pub-list-item" style="margin-bottom: 1rem" itemscope itemtype="http://schema.org/CreativeWork">
<span itemprop="author">
J. Fränken, E. Zelikman, R. Rafailov, K. Gandhi, T. Gerstenberg, N. D. Goodman</span>

(2024).

<a href="https://cicl.stanford.edu/publication/franken2024sami/" itemprop="name">Self-supervised alignment with mutual information: Learning to follow principles without preference labels</a>.
<em>Advances in Neural Information Processing Systems</em>.




<p>




<a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2404.14313" target="_blank" rel="noopener">
Preprint
</a>


<a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://cicl.stanford.edu/papers/franken2024sami.pdf" target="_blank" rel="noopener">
PDF
</a>














<a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/janphilippfranken/sami" target="_blank" rel="noopener">
Github
</a>


</p>

</div>
<div class="pub-list-item" style="margin-bottom: 1rem" itemscope itemtype="http://schema.org/CreativeWork">
<span itemprop="author">
E. Jin, Z. Huang, J. Fränken, W. Liu, H. Cha, E. Brockbank, S. Wu, R. Zhang, J. Wu, T. Gerstenberg</span>

Expand Down Expand Up @@ -505,7 +552,7 @@ <h2 id="publications">Publications</h2>



<a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/pdf/2406.15917" target="_blank" rel="noopener">
<a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2406.15917" target="_blank" rel="noopener">
Preprint
</a>

Expand Down
2 changes: 1 addition & 1 deletion docs/publication/du2024robotic/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -315,7 +315,7 @@ <h3>Abstract</h3>



<a class="btn btn-outline-primary my-1 mr-1" href="https://arxiv.org/pdf/2406.15917" target="_blank" rel="noopener">
<a class="btn btn-outline-primary my-1 mr-1" href="https://arxiv.org/abs/2406.15917" target="_blank" rel="noopener">
Preprint
</a>

Expand Down
4 changes: 2 additions & 2 deletions docs/publication/franken2024sami/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -111,9 +111,9 @@
<meta property="og:description" content="When prompting a language model (LM), users frequently expect the model to adhere to a set of behavioral principles across diverse tasks, such as producing insightful content while avoiding harmful or biased language. Instilling such principles into a model can be resource-intensive and technically challenging, generally requiring human preference labels or examples. We introduce SAMI, a method for teaching a pretrained LM to follow behavioral principles that does not require any preference labels or demonstrations. SAMI is an iterative algorithm that finetunes a pretrained LM to increase the conditional mutual information between constitutions and self-generated responses given queries from a datasest. On single-turn dialogue and summarization, a SAMI-trained mistral-7b outperforms the initial pretrained model, with win rates between 66% and 77%. Strikingly, it also surpasses an instruction-finetuned baseline (mistral-7b-instruct) with win rates between 55% and 57% on single-turn dialogue. SAMI requires a &#39;principle writer&#39; model; to avoid dependence on stronger models, we further evaluate aligning a strong pretrained model (mixtral-8x7b) using constitutions written by a weak instruction-finetuned model (mistral-7b-instruct). The SAMI-trained mixtral-8x7b outperforms both the initial model and the instruction-finetuned model, achieving a 65% win rate on summarization. Our results indicate that a pretrained LM can learn to follow constitutions without using preference labels, demonstrations, or human oversight.">
<meta property="og:locale" content="en-us">

<meta property="article:published_time" content="2024-04-22T00:00:00&#43;00:00">
<meta property="article:published_time" content="2024-10-08T00:00:00&#43;00:00">

<meta property="article:modified_time" content="2024-04-22T00:00:00&#43;00:00">
<meta property="article:modified_time" content="2024-10-08T00:00:00&#43;00:00">



Expand Down
74 changes: 73 additions & 1 deletion docs/publication/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1597,6 +1597,19 @@ <h1>Publications</h1>



















Expand Down Expand Up @@ -1733,6 +1746,65 @@ <h1>Publications</h1>



<div class='grid-sizer col-md-12 isotope-item pubtype-3 year-2024 author-'>

<div class="pub-list-item" style="margin-bottom: 1rem" itemscope itemtype="http://schema.org/CreativeWork">
<span itemprop="author">
J. Fränken, E. Zelikman, R. Rafailov, K. Gandhi, T. Gerstenberg, N. D. Goodman</span>

(2024).

<a href="https://cicl.stanford.edu/publication/franken2024sami/" itemprop="name">Self-supervised alignment with mutual information: Learning to follow principles without preference labels</a>.
<em>Advances in Neural Information Processing Systems</em>.




<p>




<a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2404.14313" target="_blank" rel="noopener">
Preprint
</a>


<a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://cicl.stanford.edu/papers/franken2024sami.pdf" target="_blank" rel="noopener">
PDF
</a>














<a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/janphilippfranken/sami" target="_blank" rel="noopener">
Github
</a>


</p>

</div>


</div>







<div class='grid-sizer col-md-12 isotope-item pubtype-3 year-2024 author-'>

<div class="pub-list-item" style="margin-bottom: 1rem" itemscope itemtype="http://schema.org/CreativeWork">
Expand Down Expand Up @@ -1921,7 +1993,7 @@ <h1>Publications</h1>



<a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/pdf/2406.15917" target="_blank" rel="noopener">
<a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2406.15917" target="_blank" rel="noopener">
Preprint
</a>

Expand Down
9 changes: 9 additions & 0 deletions docs/publication/index.xml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,15 @@
<atom:link href="https://cicl.stanford.edu/publication/index.xml" rel="self" type="application/rss+xml" />


<item>
<title>Self-supervised alignment with mutual information: Learning to follow principles without preference labels</title>
<link>https://cicl.stanford.edu/publication/franken2024sami/</link>
<pubDate>Tue, 08 Oct 2024 00:00:00 +0000</pubDate>

<guid>https://cicl.stanford.edu/publication/franken2024sami/</guid>
<description></description>
</item>

<item>
<title>MARPLE: A Benchmark for Long-Horizon Inference</title>
<link>https://cicl.stanford.edu/publication/jin2024marple/</link>
Expand Down
20 changes: 10 additions & 10 deletions docs/publication_types/3/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -111,7 +111,7 @@
<meta property="og:description" content="">
<meta property="og:locale" content="en-us">

<meta property="og:updated_time" content="2024-10-04T00:00:00&#43;00:00">
<meta property="og:updated_time" content="2024-10-08T00:00:00&#43;00:00">



Expand Down Expand Up @@ -238,6 +238,15 @@ <h1 class="pt-3">3</h1>



<div>
<h2><a href="https://cicl.stanford.edu/publication/franken2024sami/">Self-supervised alignment with mutual information: Learning to follow principles without preference labels</a></h2>
<div class="article-style">

When prompting a language model (LM), users frequently expect the model to adhere to a set of behavioral principles across diverse tasks, such as producing insightful content while avoiding harmful or biased language. Instilling such principles into …

</div>
</div>

<div>
<h2><a href="https://cicl.stanford.edu/publication/jin2024marple/">MARPLE: A Benchmark for Long-Horizon Inference</a></h2>
<div class="article-style">
Expand Down Expand Up @@ -319,15 +328,6 @@ <h2><a href="https://cicl.stanford.edu/publication/franken2024rails/">Procedural
</div>
</div>

<div>
<h2><a href="https://cicl.stanford.edu/publication/kirfel2023anticipating/">Anticipating the risks and benefits of counterfactual world simulation models</a></h2>
<div class="article-style">

This paper examines the transformative potential of Counterfactual World Simulation Models (CWSMs). CWSMs use pieces of multi-modal evidence, such as the CCTV footage or sound recordings of a road accident, to build a high-fidelity 3D reconstruction …

</div>
</div>



<nav>
Expand Down
11 changes: 10 additions & 1 deletion docs/publication_types/3/index.xml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -7,11 +7,20 @@
<generator>Hugo -- gohugo.io</generator>
<language>en-us</language>
<copyright>&amp;copy; 2024 Tobias Gerstenberg</copyright>
<lastBuildDate>Fri, 04 Oct 2024 00:00:00 +0000</lastBuildDate>
<lastBuildDate>Tue, 08 Oct 2024 00:00:00 +0000</lastBuildDate>

<atom:link href="https://cicl.stanford.edu/publication_types/3/index.xml" rel="self" type="application/rss+xml" />


<item>
<title>Self-supervised alignment with mutual information: Learning to follow principles without preference labels</title>
<link>https://cicl.stanford.edu/publication/franken2024sami/</link>
<pubDate>Tue, 08 Oct 2024 00:00:00 +0000</pubDate>

<guid>https://cicl.stanford.edu/publication/franken2024sami/</guid>
<description></description>
</item>

<item>
<title>MARPLE: A Benchmark for Long-Horizon Inference</title>
<link>https://cicl.stanford.edu/publication/jin2024marple/</link>
Expand Down
20 changes: 10 additions & 10 deletions docs/publication_types/3/page/2/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -111,7 +111,7 @@
<meta property="og:description" content="">
<meta property="og:locale" content="en-us">

<meta property="og:updated_time" content="2024-10-04T00:00:00&#43;00:00">
<meta property="og:updated_time" content="2024-10-08T00:00:00&#43;00:00">



Expand Down Expand Up @@ -238,6 +238,15 @@ <h1 class="pt-3">3</h1>



<div>
<h2><a href="https://cicl.stanford.edu/publication/kirfel2023anticipating/">Anticipating the risks and benefits of counterfactual world simulation models</a></h2>
<div class="article-style">

This paper examines the transformative potential of Counterfactual World Simulation Models (CWSMs). CWSMs use pieces of multi-modal evidence, such as the CCTV footage or sound recordings of a road accident, to build a high-fidelity 3D reconstruction …

</div>
</div>

<div>
<h2><a href="https://cicl.stanford.edu/publication/franken2023rails/">Off The Rails: Procedural Dilemma Generation for Moral Reasoning</a></h2>
<div class="article-style">
Expand Down Expand Up @@ -319,15 +328,6 @@ <h2><a href="https://cicl.stanford.edu/publication/zhang2023llm/">You are what y
</div>
</div>

<div>
<h2><a href="https://cicl.stanford.edu/publication/cao2023semantics/">A Semantics for Causing, Enabling, and Preventing Verbs Using Structural Causal Models</a></h2>
<div class="article-style">

When choosing how to describe what happened, we have a number of causal verbs at our disposal. In this paper, we develop a model-theoretic formal semantics for nine causal verbs that span the categories of CAUSE, ENABLE, and PREVENT. We use …

</div>
</div>



<nav>
Expand Down
Loading

0 comments on commit bd32ed5

Please sign in to comment.