{"id":10427,"date":"2026-02-12T07:02:27","date_gmt":"2026-02-12T07:02:27","guid":{"rendered":"https:\/\/mailitics.com\/index.php\/2026\/02\/12\/2602-10273\/"},"modified":"2026-02-12T07:02:27","modified_gmt":"2026-02-12T07:02:27","slug":"2602-10273","status":"publish","type":"post","link":"https:\/\/mailitics.com\/index.php\/2026\/02\/12\/2602-10273\/","title":{"rendered":"Power-SMC: Low-Latency Sequence-Level Power Sampling for Training-Free LLM Reasoning"},"content":{"rendered":"<p>    Power-SMC: Low-Latency Sequence-Level Power Sampling for Training-Free LLM Reasoning<br \/>\n \t<BR><br \/>\n<BR><\/BR><br \/>\n    <!-- no image --><br \/>\n \t<BR><br \/>\n<BR><\/BR><\/p>\n<div>arXiv:2602.10273v1 Announce Type: new<br \/>\nAbstract: Many recent reasoning gains in large language models can be explained as distribution sharpening: biasing generation toward high-likelihood trajectories already supported by the pretrained model, rather than modifying its weights. A natural formalization is the sequence-level power distribution $pi_alpha(ymid x)propto p_theta(ymid x)^alpha$ ($alpha&gt;1$), which concentrates mass on whole sequences instead of adjusting token-level temperature. Prior work shows that Metropolis&#8211;Hastings (MH) sampling from this distribution recovers strong reasoning performance, but at order-of-magnitude inference slowdowns. We introduce Power-SMC, a training-free Sequential Monte Carlo scheme that targets the same objective while remaining close to standard decoding latency. Power-SMC advances a small particle set in parallel, corrects importance weights token-by-token, and resamples when necessary, all within a single GPU-friendly batched decode. We prove that temperature $tau=1\/alpha$ is the unique prefix-only proposal minimizing incremental weight variance, interpret residual instability via prefix-conditioned R&#8217;enyi entropies, and introduce an exponent-bridging schedule that improves particle stability without altering the target. On MATH500, Power-SMC matches or exceeds MH power sampling while reducing latency from $16$&#8211;$28times$ to $1.4$&#8211;$3.3times$ over baseline decoding.<\/div>\n<p> \t<BR><br \/>\n <BR><\/BR><br \/>\n    Seyedarmin Azizi, Erfan Baghaei Potraghloo, Minoo Ahmadi, Souvik Kundu, Massoud Pedram<br \/>\n \t<BR><br \/>\n<BR><\/BR><br \/>\n<a href=\"https:\/\/arxiv.org\/abs\/2602.10273\">Go to original source<\/a><br \/>\n \t<BR><br \/>\n <BR><\/BR><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Power-SMC: Low-Latency Sequence-Level Power Sampling for Training-Free LLM Reasoning arXiv:2602.10273v1 Announce Type: new Abstract: Many recent reasoning gains in large language models can be explained as distribution sharpening: biasing generation toward high-likelihood trajectories already supported by the pretrained model, rather than modifying its weights. A natural formalization is the sequence-level power distribution $pi_alpha(ymid x)propto p_theta(ymid [&hellip;]<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[62,113,112],"tags":[4766,122,2089],"class_list":["post-10427","post","type-post","status-publish","format-standard","hentry","category-aimldsaimlds","category-cs-lg","category-stat-ml","tag-latency","tag-power","tag-smc"],"_links":{"self":[{"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/posts\/10427"}],"collection":[{"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/comments?post=10427"}],"version-history":[{"count":0,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/posts\/10427\/revisions"}],"wp:attachment":[{"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/media?parent=10427"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/categories?post=10427"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/tags?post=10427"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}