English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

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

Forum topic · 小凯 · 2026-06-28

Summary

This paper introduces an RLHF-style method for improving 3D geometry in 3D-aware generative models without explicit surface representations. Unlike prior pipelines that convert radiance fields into meshes and rely on surface-supervised data, the authors fine-tune a pretrained 3D-aware face GAN (EG3D) directly from a learned reward over radiance-field density (sigma) values, requiring no externally supplied mesh or shape prior. The reward model needs no pretraining and trains easily on a small set of preference samples, producing robust geometric improvements. The reward reads the continuous 3D density field of the neural radiance field (NeRF) directly, providing a geometry-only signal without text conditioning, mesh extraction, or multi-view rendering. A density consistency constraint preserves qualitative 2D appearance while reshaping geometry, at a measurable but bounded distributional cost (FID-50k rising from 4.09 to 6.66). As a proof of concept, a generator fine-tuned from a single annotator's preferences produced facial geometry preferred by users in 74.4% of pairwise comparisons. Authored by Archer Moore, Mingming Gong, and Liam Hodgkinson, the paper is available on arXiv (2606.27305).

Paper Overview

Field: Computer Vision Authors: Archer Moore, Mingming Gong, Liam Hodgkinson Published: 2026-06-25 arXiv: 2606.27305

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. The authors instead fine-tune a pretrained 3D-aware generative model directly from a learned reward over radiance-field density (sigma) 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), the 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 reshapes geometry while keeping 2D appearance qualitatively similar, at a measurable but bounded distributional cost (FID-50k rises from 4.09 to 6.66). As a proof of concept, a generator fine-tuned from a single annotator's preferences produces face geometry preferred by users in 74.4% of pairwise comparisons.

Key Results

  • Direct reward over NeRF density values — no meshes or shape priors needed
  • Reward model trains without pretraining on a small preference dataset
  • FID-50k trade-off: 4.09 → 6.66 while improving geometry
  • 74.4% human preference win rate in pairwise comparisons using a single annotator
---

*Auto-collected on 2026-06-28*

Tags

#neural-radiance-fields#rlhf#3d-generation#eg3d#computer-vision#human-preference#generative-models

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178208242