Hierarchical Group-Conditional Conformal Risk Control for Selective Prediction in Language Models
2026-07-27 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors address a problem where large language models can make too many errors on certain subgroups, even if the overall error rate seems fine. They propose HG-CRC, a new method that ensures error limits hold true across different user-defined groups by using a hierarchy and careful statistical adjustments. Their approach works without retraining the model and is tested on several language models and datasets, where it almost completely prevents errors violating the set risk budget. The results vary by benchmark, but overall HG-CRC improves fairness and reliability in model predictions.
large language modelsconformal risk controlselective predictionhierarchical calibrationBonferroni correctionrisk guaranteesgroup fairnesspost-hoc calibrationabstentionmodel evaluation
Authors
Murilo Salem, Luísa Böhm, Daniel Pontes, Anderson Ferrugem
Abstract
Large language models serve heterogeneous populations structured by domain, topic difficulty, and linguistic style. Conformal risk control (CRC) gives rigorous marginal risk guarantees for selective prediction with abstention, but marginal guarantees do not imply per-group ones: a model can meet the population budget while systematically over-exposing subgroups to errors. Under mild shift in group composition, standard CRC violates the budget in up to 47% of trials. We propose HG-CRC (Hierarchical Group-Conditional CRC), a post-hoc calibration framework enforcing simultaneous risk guarantees across all nodes of a user-defined group hierarchy. It applies a Bonferroni correction over nodes and a leaf-first policy that uses the most specific applicable threshold, falling back to coarser nodes when a finer one is uncertified or rejects the example. It needs only a held-out calibration set, with no retraining. We evaluate on three models (Qwen3-4B, Llama-3.1-8B-Instruct, Gemma-3-4B) and two benchmarks (ARC Challenge, MMLU-Pro) across eight configurations probing IID generalization, heterogeneity, mixture/domain/prompt/difficulty shift, label noise, and quantization. Main result: HG-CRC reaches an empirical 0% violation rate and WGER=0 on ARC Challenge for high-accuracy models (Qwen3-4B, Llama-3.1-8B). At 500 bootstrap trials these zeros are empirical upper bounds (true rate up to 0.6%), not certified. Results are benchmark-specific: on MMLU-Pro these models abstain entirely or (Llama) retain WGER=0.014. Gemma-3-4B, poorly calibrated here, degrades gracefully by abstaining. Participation cost vs. global CRC is 22 to 37 points. Ablations show hierarchical depth clears the budget: removing difficulty level returns violations to about 11%. Bonferroni is needed for the theoretical guarantee, though its empirical effect matters only with many nodes.