<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Coling on Fahim Dalvi</title>
    <link>https://fdalvi.github.io/tags/coling/</link>
    <description>Recent content in Coling on Fahim Dalvi</description>
    <generator>Hugo</generator>
    <language>en</language>
    <lastBuildDate>Mon, 20 Jan 2025 13:00:00 +0300</lastBuildDate>
    <atom:link href="https://fdalvi.github.io/tags/coling/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Paper Accepted at COLING 2025</title>
      <link>https://fdalvi.github.io/blog/2025-01-20-aradice-benchmarks-for-arabic-dialects/</link>
      <pubDate>Mon, 20 Jan 2025 13:00:00 +0300</pubDate>
      <guid>https://fdalvi.github.io/blog/2025-01-20-aradice-benchmarks-for-arabic-dialects/</guid>
      <description>&lt;p&gt;Excited to share that our paper &lt;a href=&#34;https://aclanthology.org/2025.coling-main.283/&#34;&gt;AraDiCE: Benchmarks for Dialectal and Cultural Capabilities in LLMs&lt;/a&gt; has been accepted at &lt;a href=&#34;https://coling2025.org/&#34;&gt;COLING 2025&lt;/a&gt; in Abu Dhabi.&lt;/p&gt;&#xA;&lt;p&gt;Arabic isn&amp;rsquo;t just one language — it&amp;rsquo;s a family of dialects that vary dramatically from region to region. Yet most LLM evaluations treat Arabic as a monolith, using only Modern Standard Arabic (MSA). This paper addresses that gap by introducing &lt;strong&gt;AraDiCE&lt;/strong&gt;, a benchmark that evaluates LLMs on both dialectal and cultural dimensions. We evaluated several LLMs on these benchmarks and found an interesting pattern: Arabic-specific models like Fanar, Jais and AceGPT do outperform general multilingual models on dialectal tasks. But significant challenges remain — particularly in dialect identification, generation, and translation.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
