Sculpting NeRF Geometry: Human-Preference Fine-Tuning of a 3D-Aware Face GAN

2026-06-25

AI Summary

This work addresses the challenge that existing 3D generation methods often rely on explicit surface representations or extensive shape supervision, making direct optimization of neural radiance field (NeRF) geometry difficult. The authors propose a novel approach that requires neither meshes, text prompts, nor multi-view supervision. Instead, they fine-tune a pretrained 3D-aware face GAN (EG3D) using human preference feedback via a lightweight reward model, optimizing only the NeRF density field to enhance geometric quality while introducing a density consistency constraint to preserve appearance fidelity. Experimental results demonstrate that the fine-tuned model generates facial geometries preferred by users in 74.4% of pairwise comparisons, with only a modest increase in FID-50k from 4.09 to 6.66, effectively balancing geometric improvement and visual quality.

Abstract

Reinforcement learning from human feedback (RLHF) for 3D generation is now established across a number of works, but most existing pipelines optimise explicit surface representations, often by converting radiance fields into meshes and training heavily on surface-supervised data. We instead fine-tune a pretrained 3D-aware generative model directly from a learned reward over radiance-field density ($σ$) values, with no externally supplied mesh or shape prior. The reward model requires no pretraining, trains easily on a small set of preference samples, and yields robust improvement in 3D geometry. Working on an unconditional 3D-aware face GAN (EG3D), our reward reads the continuous 3D density field of the neural radiance field (NeRF) directly and supplies a geometry-only learning signal, requiring neither text conditioning, mesh extraction, nor multi-view rendering. A density-consistency constraint keeps the 2D appearance qualitatively similar while the geometry is reshaped, at a measurable but bounded distributional cost (FID-50k rises from 4.09 to 6.66): the fine-tuned generator, trained from the preferences of a single annotator as a proof of concept, produces face geometries preferred by users in 74.4% of pairwise comparisons.

Problem

• 3D-aware GAN
• Neural Radiance Fields
• human preference
• geometry optimization
• reinforcement learning from human feedback

Innovation

• RLHF
• NeRF
• 3D-aware GAN
• density field
• geometry optimization