<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>NUS AI Research Day | Web IR NLP Group @ NUS</title><link>https://wing.comp.nus.edu.sg/tag/nus-ai-research-day/</link><atom:link href="https://wing.comp.nus.edu.sg/tag/nus-ai-research-day/index.xml" rel="self" type="application/rss+xml"/><description>NUS AI Research Day</description><generator>Hugo Blox Builder (https://hugoblox.com)</generator><language>en-us</language><lastBuildDate>Tue, 22 Sep 2026 11:00:00 +0800</lastBuildDate><image><url>https://wing.comp.nus.edu.sg/media/icon_hu_ce731d16d7c8719.png</url><title>NUS AI Research Day</title><link>https://wing.comp.nus.edu.sg/tag/nus-ai-research-day/</link></image><item><title>Sahajpreet Singh Wins Best Poster Presenter Award at NUS AI Research Day 2026</title><link>https://wing.comp.nus.edu.sg/event/26-09-22-nus-ai-research-day/</link><pubDate>Tue, 22 Sep 2026 11:00:00 +0800</pubDate><guid>https://wing.comp.nus.edu.sg/event/26-09-22-nus-ai-research-day/</guid><description>&lt;p&gt;At NUS AI Research Day 2026, &lt;strong&gt;&lt;a href="https://wing.comp.nus.edu.sg/author/sahajpreet-singh/" target="_blank" rel="noopener"&gt;Sahajpreet Singh&lt;/a&gt;&lt;/strong&gt; presented &lt;em&gt;Camellia: Benchmarking Cultural Biases in LLMs for Asian Languages&lt;/em&gt; and received the &lt;strong&gt;Best Poster Presenter Award&lt;/strong&gt;. The university-wide event brought together AI researchers and participants to share their work; WING member &lt;a href="https://wing.comp.nus.edu.sg/author/yisong-miao/" target="_blank" rel="noopener"&gt;Yisong Miao&lt;/a&gt; served on the programme committee.&lt;/p&gt;
&lt;p&gt;Camellia is a benchmark for studying how large language models handle cultural context across nine Asian languages and six cultures. It contains 19,530 manually annotated entities and 2,173 masked social-media contexts. The paper evaluates models on cultural context adaptation, sentiment association, and entity extractive question answering. Its results show that models can struggle to adapt to cultural context and that performance varies across models and languages.&lt;/p&gt;
&lt;p&gt;The research was presented on &lt;strong&gt;Tuesday, 22 September 2026&lt;/strong&gt;, from &lt;strong&gt;11:00 am to 4:00 pm&lt;/strong&gt; at &lt;strong&gt;COM3 Multi-purpose Hall (#01-27), 11 Research Link, Singapore 119391&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Paper:&lt;/strong&gt; &lt;a href="https://arxiv.org/abs/2510.05291" target="_blank" rel="noopener"&gt;Camellia: Benchmarking Cultural Biases in LLMs for Asian Languages&lt;/a&gt; (accepted at EMNLP 2026 Main Conference).&lt;/p&gt;
&lt;p&gt;
&lt;figure &gt;
&lt;div class="d-flex justify-content-center"&gt;
&lt;div class="w-100" &gt;&lt;img alt="WING members and collaborators at NUS AI Research Day, with the Camellia poster on display." srcset="
/event/26-09-22-nus-ai-research-day/featured_hu_fa655e19c77aaab.webp 400w,
/event/26-09-22-nus-ai-research-day/featured_hu_e608d0816610f501.webp 760w,
/event/26-09-22-nus-ai-research-day/featured_hu_8120643157536d89.webp 1200w"
src="https://wing.comp.nus.edu.sg/event/26-09-22-nus-ai-research-day/featured_hu_fa655e19c77aaab.webp"
width="760"
height="533"
loading="lazy" data-zoomable /&gt;&lt;/div&gt;
&lt;/div&gt;&lt;/figure&gt;
&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Based on &lt;a href="https://x.com/wing_nus/status/2117185961406632044" target="_blank" rel="noopener"&gt;WING&amp;rsquo;s event post&lt;/a&gt; and &lt;a href="https://x.com/YisongMiao/status/2117188135100760173" target="_blank" rel="noopener"&gt;Yisong Miao&amp;rsquo;s post&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;</description></item></channel></rss>