Sampling More, Getting Less: Calibration is the Diversity Bottleneck in LLMs
For researchers and practitioners using LLMs for diverse generation, this paper provides a causal explanation for diversity collapse, shifting focus from sampling methods to model calibration.
The paper identifies that diversity collapse in LLMs is caused by two forms of miscalibration in the model's probability distribution: order calibration (valid tokens not reliably ranked above invalid ones) and shape calibration (probability mass overly concentrated on few valid continuations). Across 14 models, they show this is a fundamental distributional issue, not just a sampling heuristic limitation.
Diversity is essential for language-model applications ranging from creative generation to scientific discovery, yet modern LLMs often collapse into a narrow subset of plausible outputs. While prior work has developed benchmarks for measuring this lack of diversity, less is known about how the step-by-step probability distributions at inference time cause the problem. We introduce a validity--diversity framework that attributes diversity collapse to how an LLM allocates probability mass across valid and invalid continuations during decoding. This framework decomposes the bottleneck into two complementary forms of miscalibration. First, order calibration: valid tokens are not reliably ranked above invalid tokens, so rank-based cutoff rules must trade off between recovering valid continuations and admitting invalid ones. Second, shape calibration: probability mass is overly concentrated only on few valid continuations while having a heavy-tail of mixed valid and invalid tokens, so maintaining high validity limits diversity. We formalize both mechanisms and show that local failures compound across decoding steps, producing strong sequence-level losses in diversity. Empirically, we develop controlled diagnostics for probing these bottlenecks, including tasks with exactly known valid sets and oracle cutoff baselines. Across 14 language models spanning multiple families and scales, we find that diversity collapse is not merely a limitation of particular sampling heuristics, but a consequence of order and shape miscalibration in the LLM distribution.