# Conformal Fair Decoding for Large Language Model Recommenders

DevFeed: [Conformal Fair Decoding for Large Language Model Recommenders](<https://devfeed.tech/articles/conformal-fair-decoding-for-large-language-model-recommenders-125912.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.10285>)

Author: Ziwei Wu, Jingrui He

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [recommendation systems](<https://devfeed.tech/topics/recommendation-systems.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [ai-governance](<https://devfeed.tech/topics/ai-governance.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>)

Tags: [amazon](<https://devfeed.tech/tags/amazon.md>), [candidate](<https://devfeed.tech/tags/candidate.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [decoding](<https://devfeed.tech/tags/decoding.md>), [fairness](<https://devfeed.tech/tags/fairness.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [llama3](<https://devfeed.tech/tags/llama3.md>), [qwen3](<https://devfeed.tech/tags/qwen3.md>), [recommendation](<https://devfeed.tech/tags/recommendation.md>), [semantic](<https://devfeed.tech/tags/semantic.md>), [zero-shot](<https://devfeed.tech/tags/zero-shot.md>)

## AI overview

Fair-Rec controls counterfactual semantic disparity in LLM recommendation lists when model weights and full-vocabulary logits are unavailable. It uses an offline-calibrated disparity threshold: a hard decoder filters candidates against it, while a soft decoder balances candidate likelihood and disparity and adjusts a penalty based on observed violations. Under a stated uniform enforceability assumption, the authors prove a pathwise O(1/T) upper bound on the soft decoder's excess empirical violation rate. In experiments on MovieLens-1M and Amazon Movies & TV with Qwen3-8B and Llama3.1-8B-Instruct, the soft variant reduced measured disparity versus zero-shot recommendation, UP5, and FACTER while maintaining competitive HR@10 and NDCG@10.

## Source excerpt

arXiv:2611.10285v1 Announce Type: new Abstract: Large language models (LLMs) can generate personalized recommendations without task-specific training, but changing a protected attribute while holding interaction history fixed can alter the semantic content of recommended items. We study fairness control when model weights and full-vocabulary logits are unavailable. We propose Fair- Rec, an item-level decoding framework that uses an offline-calibrated semantic-disparity threshold and controls candidate selection as recommendation lists are generated. A hard variant filters candidates against the threshold; a soft variant trades off candidate likelihood and disparity while adapting a penalty using observed violations. Under an explicit uniform enforceability assumption, we establish a pathwise O(1/T) upper bound on the excess empirical violation rate of the soft decoder. Experiments on MovieLens-1M and Amazon Movies & TV using Qwen3-8B and Llama3.1-8B-Instruct show that the soft variant lowers measured counterfactual disparity relative to zero- =shot recommendation, UP5, and FACTER, while retaining competitive HR@10 and NDCG@10.