{"id":6878,"date":"2025-09-16T07:03:43","date_gmt":"2025-09-16T07:03:43","guid":{"rendered":"https:\/\/mailitics.com\/index.php\/2025\/09\/16\/2509-11208\/"},"modified":"2025-09-16T07:03:43","modified_gmt":"2025-09-16T07:03:43","slug":"2509-11208","status":"publish","type":"post","link":"https:\/\/mailitics.com\/index.php\/2025\/09\/16\/2509-11208\/","title":{"rendered":"Predictable Compression Failures: Why Language Models Actually Hallucinate"},"content":{"rendered":"

Predictable Compression Failures: Why Language Models Actually Hallucinate
\n \t

\n
<\/BR>
\n
\n \t

\n
<\/BR><\/p>\n

arXiv:2509.11208v1 Announce Type: new
\nAbstract: Large language models perform near-Bayesian inference yet violate permutation invariance on exchangeable data. We resolve this by showing transformers minimize expected conditional description length (cross-entropy) over orderings, $mathbb{E}_pi[ell(Y mid Gamma_pi(X))]$, which admits a Kolmogorov-complexity interpretation up to additive constants, rather than the permutation-invariant description length $ell(Y mid X)$. This makes them Bayesian in expectation, not in realization. We derive (i) a Quantified Martingale Violation bound showing order-induced deviations scale as $O(log n)$ with constants; (ii) the Expectation-level Decompression Law linking information budgets to reliability for Bernoulli predicates; and (iii) deployable planners (B2T\/RoH\/ISR) for answer\/abstain decisions. Empirically, permutation dispersion follows $a+bln n$ (Qwen2-7B $b approx 0.377$, Llama-3.1-8B $b approx 0.147$); permutation mixtures improve ground-truth likelihood\/accuracy; and randomized dose-response shows hallucinations drop by $sim 0.13$ per additional nat. A pre-specified audit with a fixed ISR=1.0 achieves near-0% hallucinations via calibrated refusal at 24% abstention. The framework turns hallucinations into predictable compression failures and enables principled information budgeting.<\/div>\n

\t

\n
<\/BR>
\n Leon Chlon, Ahmed Karim, Maggie Chlon
\n \t

\n
<\/BR>
\nGo to original source<\/a>
\n \t

\n
<\/BR><\/p>\n","protected":false},"excerpt":{"rendered":"

Predictable Compression Failures: Why Language Models Actually Hallucinate arXiv:2509.11208v1 Announce Type: new Abstract: Large language models perform near-Bayesian inference yet violate permutation invariance on exchangeable data. We resolve this by showing transformers minimize expected conditional description length (cross-entropy) over orderings, $mathbb{E}_pi[ell(Y mid Gamma_pi(X))]$, which admits a Kolmogorov-complexity interpretation up to additive constants, rather than the […]<\/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":[2831,1993,3794],"class_list":["post-6878","post","type-post","status-publish","format-standard","hentry","category-aimldsaimlds","category-cs-lg","category-stat-ml","tag-compression","tag-permutation","tag-predictable"],"_links":{"self":[{"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/posts\/6878"}],"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=6878"}],"version-history":[{"count":0,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/posts\/6878\/revisions"}],"wp:attachment":[{"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/media?parent=6878"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/categories?post=6878"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/tags?post=6878"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}