New Delegation-Based Method Outperforms Majority Voting for Multi-Sample LLM Inference
Researchers introduced Propagational Proxy Voting (PPV), an unsupervised aggregation method that beats majority voting on the MMLU-Pro benchmark by 1.5 percentage points overall and 2.24 points on harder questions. PPV leverages two signals that majority voting discards: letter entropy within answer groups and geometric coherence of reasoning embeddings between groups. The method requires no training data or gold labels, making it practical for improving LLM accuracy in real-world applications.
A new arXiv paper proposes Propagational Proxy Voting (PPV), an unsupervised aggregation method for combining multiple LLM-generated answers that outperforms standard majority voting. The method works by partitioning 128 sampled generations into 16 groups and analyzing two signals: within-group letter-level semantic entropy and between-group reasoning embedding geometry. PPV uses these signals to construct a stochastic delegation matrix that determines how much weight each voter keeps on its own answer (WHEN) and how it distributes remaining weight across peers (WHOM). Testing on MMLU-Pro showed improvements of +1.5 percentage points overall and +2.24 points on non-trivial questions, with statistical significance (paired McNemar p ~ 1.0e-14, n = 8,099). The authors demonstrate cases where PPV correctly identifies that a minority cluster with geometrically coherent reasoning should override a larger but incoherent majority cluster, and report negative results showing that confidence-based ensemble methods cannot close the gap to oracle performance.
What's missing
The paper does not discuss computational overhead or runtime comparisons between PPV and majority voting, nor does it evaluate performance on benchmarks beyond MMLU-Pro or on different types of LLM tasks (e.g., generation, reasoning, factual recall). The generalizability of the method to other domains and model sizes remains unclear.
What different sources said
- arXiv cs.AICenter
When Does Delegation Beat Majority? A Delegation-Based Aggregator for Multi-Sample LLM Inference
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.