Paul Houle<p>-𝟘: The Power of Negative Zero: Datatype Customization for Quantized Large Language Models</p><p>(... they get an amazing boost in performance from this!)</p><p><a href="https://arxiv.org/abs/2501.04052" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="">arxiv.org/abs/2501.04052</span><span class="invisible"></span></a></p><p><a href="https://mastodon.social/tags/ai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ai</span></a> <a href="https://mastodon.social/tags/ml" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ml</span></a> <a href="https://mastodon.social/tags/cs" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>cs</span></a> <a href="https://mastodon.social/tags/neuralnet" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>neuralnet</span></a> <a href="https://mastodon.social/tags/math" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>math</span></a> <a href="https://mastodon.social/tags/efficiency" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>efficiency</span></a> <a href="https://mastodon.social/tags/goodnews" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>goodnews</span></a></p>