Research map / Image generation / Autoregressive and latent generation
FlashMesh: Faster and Better Autoregressive Mesh Synthesis via Structured Speculation
Tingrui Shen, Yiheng Zhang, Chen Tang, Chuan Ping, Zixing Zhao, Le Wan, Yuwang Wang, Ronggang Wang, Shengfeng He
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add SP-Block only (coordinate-level speculation) (ablation) | Meshtron 1B · Mesh generation ablation (Meshtron 1B) | Tokens per second: 95.5 -> 109.7 | Improved |
| add HF-Block to SP-Block (ablation) | FlashMesh (Meshtron 1B) · Mesh generation ablation (Meshtron 1B) | Tokens per second: 109.7 -> 176.5 | Improved |
| add correction mechanism (ablation) | FlashMesh (Meshtron 1B) · Mesh generation ablation (Meshtron 1B) | Tokens per second: 176.5 -> 180.4 | Improved |
| add correction mechanism (ablation) | FlashMesh (Meshtron 1B) · Mesh generation ablation (Meshtron 1B) | Chamfer Distance: 0.120 -> 0.120 | No clear effect |
| increase draft tokens from 18-18 to 27-27 (ablation) | FlashMesh (Meshtron 1B) · Draft token count (Meshtron 1B) | Tokens per second: 179.9 -> 114.4 | Hurt |
| increase draft tokens from 18-18 to 27-27 (ablation) | FlashMesh (Meshtron 1B) · Draft token count (Meshtron 1B) | Chamfer Distance: 0.120 -> 0.127 | Hurt |
| use 9-9 draft tokens instead of 18-18 (ablation) | FlashMesh (Meshtron 1B) · Draft token count (Meshtron 1B) | Tokens per second: 179.9 -> 139.9 | Hurt |
| apply FlashMesh to 0.5B backbone | Meshtron 0.5B · Model size (Meshtron backbone) | Chamfer Distance: 0.137 -> 0.140 | Hurt |
| apply FlashMesh to 2B backbone | Meshtron 2B · ShapeNetV2+gObjaverse test (1000 meshes) | Chamfer Distance: 0.092 -> 0.089 | Improved |
| apply FlashMesh to Meshtron 1B | Meshtron 1B · ShapeNetV2+gObjaverse test (1000 meshes) | Tokens per second: 98.6 -> 180.4 | Improved |
| vary loss weight gamma (0.1, 0.3, 0.5) (ablation) | FlashMesh (Meshtron 1B) · Mesh generation ablation (Meshtron 1B) | Chamfer Distance: 0.120 for all gamma | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Mesh-RFT — Mesh-RFT: Enhancing Mesh Generation via Fine-grained Reinforcement Fine-Tuning (NeurIPS 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ARMesh: Autoregressive Mesh Generation via Next-Level-of-Detail Prediction (NeurIPS 2025 · similarity 0.88)
- PivotMesh: Generic 3D Mesh Generation via Pivot Vertices Guidance (ICLR 2025 · similarity 0.88)
- TreeMeshGPT: Artistic Mesh Generation with Autoregressive Tree Sequencing (CVPR 2025 · similarity 0.87)
- MeshAnything: Artist-Created Mesh Generation with Autoregressive Transformers (ICLR 2025 · similarity 0.86)
- MeshGPT: Generating Triangle Meshes with Decoder-Only Transformers (CVPR 2024 · similarity 0.86)
- LANTERN: Accelerating Visual Autoregressive Models with Relaxed Speculative Decoding (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
