dorsal/arxiv
View SchemaHallucinations Live in Variance
| Authors | Aaron R. Flouro, Shawn P. Chadwick |
|---|---|
| Categories | |
| ArXiv ID | 2601.07058vv1 |
| URL | https://arxiv.org/abs/2601.07058 |
| License | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
Abstract
Benchmarks measure whether a model is correct. They do not measure whether a model is reliable. This distinction is largely academic for single-shot inference, but becomes critical for agentic AI systems, where a single rephrased prompt can trigger cascading failures in multi-step execution. Yet this form of instability is not captured by existing evaluations. Hallucinations live in variance: they arise when semantically equivalent prompts activate inconsistent internal pathways, producing divergent outputs. Consistent but incorrect outputs reflect bias or missing knowledge; confident guessing reflects calibration failure. Neither constitutes hallucination under this definition. When error is variance-dominated, reducing redundant pathways improves reliability without adding knowledge. We formalize this through Semantic Stability (SS), measured via Paraphrase Consistency (PC@k): generate k paraphrases, greedy decode each, compute mode agreement. SS is a diagnostic for variance-driven unreliability, not a method for improving correctness. We show that a dense Qwen3-0.6B agrees with itself only 23.8% of the time; at 32% sparsity, agreement jumps to 55.9%. A phase diagram reveals the sweet spot where variance reduction outpaces bias accumulation, and regimes where stability collapses onto wrong answers.
{
"annotation_id": "e1a785e0-d24f-47fd-b226-9c14f46babe2",
"date_created": "2026-02-17T05:53:08.878000Z",
"date_modified": "2026-02-17T05:53:08.878000Z",
"file_hash": "77db177d2b9096dc4c7cb19bdcce252d8279cbb11c8426f8c0eb998242b18308",
"private": false,
"record": {
"abstract": "Benchmarks measure whether a model is correct. They do not measure whether a model is reliable. This distinction is largely academic for single-shot inference, but becomes critical for agentic AI systems, where a single rephrased prompt can trigger cascading failures in multi-step execution. Yet this form of instability is not captured by existing evaluations.\n Hallucinations live in variance: they arise when semantically equivalent prompts activate inconsistent internal pathways, producing divergent outputs. Consistent but incorrect outputs reflect bias or missing knowledge; confident guessing reflects calibration failure. Neither constitutes hallucination under this definition. When error is variance-dominated, reducing redundant pathways improves reliability without adding knowledge. We formalize this through Semantic Stability (SS), measured via Paraphrase Consistency (PC@k): generate k paraphrases, greedy decode each, compute mode agreement. SS is a diagnostic for variance-driven unreliability, not a method for improving correctness.\n We show that a dense Qwen3-0.6B agrees with itself only 23.8% of the time; at 32% sparsity, agreement jumps to 55.9%. A phase diagram reveals the sweet spot where variance reduction outpaces bias accumulation, and regimes where stability collapses onto wrong answers.",
"arxiv_id": "2601.07058",
"authors": [
"Aaron R. Flouro",
"Shawn P. Chadwick"
],
"categories": [
"cs.LG",
"cs.AI"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"title": "Hallucinations Live in Variance",
"url": "https://arxiv.org/abs/2601.07058",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "52d70890-1ab1-4e5a-8aea-c40f7e151478",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}