<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
</head>
<body style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class="">
<div class="elementToProof" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
This is an announcement of Mansi Sakarvadia's MS Presentation</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
===============================================</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
Candidate: Mansi Sakarvadia</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
<br class="">
</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
Date: Friday, April 05, 2024</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
<br class="">
</div>
<div class="elementToProof" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
Time: 2 pm CT</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
<br class="">
</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
Remote Location: <a href="https://uchicago.zoom.us/j/9643461977?pwd=VUJhRVNiQ0lHMWdGUVVNRE9rOUdNQT09" class="">https://uchicago.zoom.us/j/9643461977?pwd=VUJhRVNiQ0lHMWdGUVVNRE9rOUdNQT09</a> Meeting ID: 964 346 1977 Passcode: 109461</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
<br class="">
</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
Location: JCL 390</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
<br class="">
</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
Title: Memory Injections: Correcting Multi-Hop Reasoning Failures during Inference in Transformer-Based Language Models</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
<br class="">
</div>
<div class="elementToProof" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
Abstract: Answering multi-hop reasoning questions requires retrieving and synthesizing information from diverse sources. Language models (LMs) struggle to perform such reasoning consistently. Here we propose an approach to pinpoint and rectify multi-hop reasoning
failures through targeted memory injections on LM attention heads. First, we analyze the per-layer activations of GPT-2 models in response to single- and multi-hop prompts. We then propose a mechanism that allows users to inject relevant prompt-specific information,
which we refer to as "memories," at critical LM locations during inference. By thus enabling the LM to incorporate additional relevant information during inference, we enhance the quality of multi-hop prompt completions. We empirically show that a simple,
efficient, and targeted memory injection into a key attention layer can often increase the probability of the desired next token in multi-hop tasks, by up to 424%. From this work, we observe that small subsets of attention heads can significantly impact the
model prediction during multi-hop reasoning. To more faithfully interpret these heads, we develop Attention Lens: an open source tool that translates the outputs of attention heads into vocabulary tokens via learned transformations called lenses. We demonstrate
the use of lenses to reveal how a model arrives at its answer and use them to localize sources of model failures such as in the case of biased and malicious language generation.</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
<br class="">
</div>
<div class="elementToProof" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
Advisors: Ian Foster and Kyle Chard</div>
<div class="" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
<br class="">
</div>
<div class="elementToProof" style="font-family: Calibri, Helvetica, sans-serif; font-size: 11pt;">
Committee Members: Ian Foster, Kyle Chard, and Ari Holtzman</div>
<div class="">
<div dir="auto" class="" style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;">
<div dir="auto" class="" style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;">
<div dir="auto" class="" style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;">
<div dir="auto" class="" style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;">
<div dir="auto" class="" style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;">
<div dir="auto" class="" style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;">
</div>
</div>
</div>
</div>
</div>
</div>
</div>
<div class="">
<div dir="auto" style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class="">
<div dir="auto" style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class="">
<div dir="auto" style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class="">
<div dir="auto" style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class="">
<div dir="auto" style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class="">
<div dir="auto" style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class="">
<br class="Apple-interchange-newline">
</div>
<br class="Apple-interchange-newline">
</div>
<br class="Apple-interchange-newline">
</div>
<br class="Apple-interchange-newline">
</div>
<br class="Apple-interchange-newline">
</div>
<br class="Apple-interchange-newline">
</div>
<br class="Apple-interchange-newline">
<br class="Apple-interchange-newline">
</div>
<br class="">
</body>
</html>