Built site for gh-pages

.nojekyll

@@ -1 +1 @@
-2ba09bbc
+7a8d4739

index.html

@@ -143,7 +143,7 @@
 
 <div class="quarto-listing quarto-listing-container-grid" id="listing-listing">
 <div class="list grid quarto-listing-cols-3">
-<div class="g-col-1" data-index="0" data-listing-date-sort="1704326400000" data-listing-file-modified-sort="1705061835129" data-listing-date-modified-sort="NaN" data-listing-reading-time-sort="25">
+<div class="g-col-1" data-index="0" data-listing-date-sort="1704326400000" data-listing-file-modified-sort="1705146805089" data-listing-date-modified-sort="NaN" data-listing-reading-time-sort="25">
 <a href="./posts/TDC2023.html" class="quarto-grid-link">
 <div class="quarto-grid-item card h-100 card-left">
 <p class="card-img-top"><img src="posts/TDC2023-sample-instances.png" style="height: 150px;"  class="thumbnail-image card-img"/></p>
@@ -166,7 +166,7 @@ <h5 class="no-anchor card-title listing-title">
 </div>
 </a>
 </div>
-<div class="g-col-1" data-index="1" data-listing-date-sort="1701302400000" data-listing-file-modified-sort="1705061835149" data-listing-date-modified-sort="NaN" data-listing-reading-time-sort="7">
+<div class="g-col-1" data-index="1" data-listing-date-sort="1701302400000" data-listing-file-modified-sort="1705146805109" data-listing-date-modified-sort="NaN" data-listing-reading-time-sort="7">
 <a href="./posts/fight_the_illusion.html" class="quarto-grid-link">
 <div class="quarto-grid-item card h-100 card-left">
 <div class="listing-item-img-placeholder card-img-top" style="height: 150px;">&nbsp;</div>
@@ -189,7 +189,7 @@ <h5 class="no-anchor card-title listing-title">
 </div>
 </a>
 </div>
-<div class="g-col-1" data-index="2" data-listing-date-sort="1687651200000" data-listing-file-modified-sort="1705061835149" data-listing-date-modified-sort="NaN" data-listing-reading-time-sort="7">
+<div class="g-col-1" data-index="2" data-listing-date-sort="1687651200000" data-listing-file-modified-sort="1705146805109" data-listing-date-modified-sort="NaN" data-listing-reading-time-sort="7">
 <a href="./posts/catalog.html" class="quarto-grid-link">
 <div class="quarto-grid-item card h-100 card-left">
 <p class="card-img-top"><img src="posts/catalog_files/figure-html/cell-9-output-1.png" style="height: 150px;"  class="thumbnail-image card-img"/></p>

posts/TDC2023.html

@@ -347,6 +347,10 @@ <h4 class="anchored" data-anchor-id="the-models-may-have-developed-internal-geom
 <p>Assume we are trying to find triggers for some payload <span class="math inline">\(s_2\)</span>. Take a completely unrelated known trigger-payload pair <span class="math inline">\((p_1, s_1)\)</span>, such that trigger <span class="math inline">\(p_1\)</span> yields a different payload <span class="math inline">\(s_1\)</span>. Then, while optimizing for payload <span class="math inline">\(s_2\)</span>, initialize the optimization at the point <span class="math inline">\(x = p_1\)</span>. This turns out to speed up the process of finding a trigger for <span class="math inline">\(s_2\)</span>, often with far fewer iterations than if we had initialized with random tokens or text from the Pile.</p>
 <p>Somehow, GCG’s first-order approximation (which it uses to select candidate mutations) is accurate enough to rapidly descend in this setting. In some cases, payload <span class="math inline">\(s_2\)</span> could be produced with <em>only 1-3 optimizer iterations</em> starting from trigger <span class="math inline">\(p_1\)</span>. We were very surprised by this. Perhaps there is a well-behaved connecting manifold that forms between the trojans? <strong>If we were to continue attempting to reverse engineer trojan insertion, understanding this phenomenon is where we would start.</strong></p>
 </section>
+<section id="section" class="level4">
+<h4 class="anchored" data-anchor-id="section">5.</h4>
+<p>For some additional details on our investigations, see <a href="https://zygi.me/blog/adventures-in-trojan-detection/#open-questions">Zygi’s personal site</a></p>
+</section>
 </section>
 <section id="red-teaming-track-takeaways" class="level1">
 <h1>Red Teaming Track Takeaways</h1>
@@ -718,7 +722,7 @@ <h4 class="anchored" data-anchor-id="tricks-that-we-found-to-improve-performance
 });
 </script>
 </div> <!-- /content -->
-<script>var lightboxQuarto = GLightbox({"selector":".lightbox","loop":true,"openEffect":"zoom","descPosition":"bottom","closeEffect":"zoom"});</script>
+<script>var lightboxQuarto = GLightbox({"selector":".lightbox","openEffect":"zoom","closeEffect":"zoom","descPosition":"bottom","loop":true});</script>
 
 
 

posts/catalog.html

@@ -814,7 +814,7 @@ <h2 class="anchored" data-anchor-id="github">GitHub</h2>
 });
 </script>
 </div> <!-- /content -->
-<script>var lightboxQuarto = GLightbox({"selector":".lightbox","loop":true,"openEffect":"zoom","closeEffect":"zoom","descPosition":"bottom"});</script>
+<script>var lightboxQuarto = GLightbox({"descPosition":"bottom","openEffect":"zoom","selector":".lightbox","closeEffect":"zoom","loop":true});</script>
 
 
 

posts/catalog.out.ipynb

@@ -297,7 +297,7 @@
         "Pythia-12B is miscalibrated on 20% of the bigrams and 45% of the\n",
         "trigrams when we ask for prediction of $p \\geq 0.45$."
       ],
-      "id": "4fbc51b1-902c-44b7-b3ba-46f4f33a1cf4"
+      "id": "393b0a4d-96e5-4817-8f6d-8aa9874713ca"
     },
     {
       "cell_type": "code",
@@ -313,7 +313,7 @@
         }
       ],
       "source": [],
-      "id": "4bbbb077-5235-4fb4-95e5-98b3cd9e12f2"
+      "id": "d6345904-eaa8-46a9-8a84-6af0a96ede58"
     },
     {
       "cell_type": "markdown",
@@ -377,7 +377,7 @@
         "The dataset is available on Huggingface:\n",
         "[pile_scan_4](https://huggingface.co/datasets/Confirm-Labs/pile_scan_4)"
       ],
-      "id": "7ac4275d-fadd-460b-a0e8-51109410d634"
+      "id": "9ac7dedc-2219-4841-994f-f0e278648347"
     },
     {
       "cell_type": "code",
@@ -391,7 +391,7 @@
         }
       ],
       "source": [],
-      "id": "46198176-f494-4cf6-8e7c-0530c5eac2a2"
+      "id": "e45be69e-cb7d-446c-8379-e30e166022b6"
     },
     {
       "cell_type": "markdown",
@@ -423,7 +423,7 @@
         "Computational Linguistics, May 2022, pp. 95–136. doi:\n",
         "[10.18653/v1/2022.bigscience-1.9](https://doi.org/10.18653/v1/2022.bigscience-1.9).</span>"
       ],
-      "id": "7019dc8f-3e50-4e00-b3ec-ad1d99950649"
+      "id": "1b4be320-18ff-4108-a568-431d308a4148"
     }
   ],
   "nbformat": 4,

sitemap.xml

@@ -2,18 +2,18 @@
 <urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
   <url>
     <loc>https://confirmlabs.org/posts/catalog.html</loc>
-    <lastmod>2024-01-12T12:17:44.093Z</lastmod>
+    <lastmod>2024-01-13T11:53:39.873Z</lastmod>
   </url>
   <url>
     <loc>https://confirmlabs.org/posts/TDC2023.html</loc>
-    <lastmod>2024-01-12T12:17:40.893Z</lastmod>
+    <lastmod>2024-01-13T11:53:36.773Z</lastmod>
   </url>
   <url>
     <loc>https://confirmlabs.org/index.html</loc>
-    <lastmod>2024-01-12T12:17:39.437Z</lastmod>
+    <lastmod>2024-01-13T11:53:35.349Z</lastmod>
   </url>
   <url>
     <loc>https://confirmlabs.org/posts/fight_the_illusion.html</loc>
-    <lastmod>2024-01-12T12:17:41.589Z</lastmod>
+    <lastmod>2024-01-13T11:53:37.433Z</lastmod>
   </url>
 </urlset>