New Delegation-Based Method Outperforms Majority Voting for Multi-Sample LLM Inference
Researchers have proposed a delegation-based aggregation method called Propagational Proxy Voting (PPV) that outperforms standard majority voting for combining multiple LLM-generated answers, gaining +1.5 percentage points on the MMLU-Pro benchmark. Majority voting, the current dominant approach, ignores two informational signals present in each sampled answer: within-group letter entropy and between-group reasoning geometry, both of which PPV explicitly leverages. The finding matters because unsupervised aggregation of multiple LLM samples is a widely used technique for improving model accuracy without additional training or labeled data.
A preprint posted to arXiv introduces Propagational Proxy Voting (PPV), an unsupervised aggregation method for multi-sample large language model (LLM) inference that surpasses majority voting on the MMLU-Pro benchmark by +1.5 percentage points overall and +2.24 percentage points on non-trivial questions, with strong statistical significance (McNemar p ≈ 1.0e-14, n = 8,099). The method works by partitioning 128 sampled model generations per question into 16 groups, computing each group's letter-level semantic entropy and reasoning embedding centroid, and feeding these into a stochastic delegation matrix whose stationary distribution determines the consensus answer. Unlike majority voting, PPV captures two signals that standard approaches discard: how confident a group is in its own answer (entropy) and how geometrically coherent its reasoning is relative to peers (embedding cosine similarity). A key illustrative example shows PPV correctly overturning a 10-to-6 majority for a wrong answer, because the majority cluster was geometrically incoherent (mean cosine -0.02) while the minority cluster was tight (+0.26). The method requires no gold labels and no auxiliary model training, making it a drop-in replacement for majority voting in existing inference pipelines. The authors also report negative results from alternative delegation strategies, helping to constrain the broader design space for unsupervised LLM aggregation.
What's missing
As a preprint, this work has not yet undergone peer review. The evaluation is limited to a single benchmark (MMLU-Pro) with one model configuration (128 samples, 16 groups), leaving open questions about generalizability across different LLMs, sample budgets, and task types. The computational overhead of PPV relative to majority voting is not fully characterized.
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.