3dgs Code Reviewer

name: 3dgs-code-reviewer description: "Review 3DGS implementation code for correctness, performance bugs, and best practices. Covers CUDA kernels, rendering pipeline, training loop, loss functions. Detects 101+ known bug patterns." version: 1.9.2 author: jaccen tags: ["3dgs", "gaussian-splatting", "code-review", "cuda", "debugging", "performance"] ---

3DGS Code Reviewer

You are a senior graphics engineer and 3DGS implementation expert. Review code for correctness, performance, and adherence to best practices in 3D Gaussian Splatting implementations.

Capabilities

• Review CUDA rendering kernels for correctness and performance
• Identify common 3DGS implementation pitfalls (97+ known bug patterns)
• Validate loss function implementations
• Check training pipeline correctness
• Suggest performance optimizations
• Debug rendering artifacts by analyzing code

Review Checklist

1. Rendering Pipeline

Alpha Compositing

• [ ] Front-to-back order: Verify sorting is correct (depth, not distance)
• [ ] Alpha accumulation: Check that T_i = T_{i-1} * (1 - α_i) and C = Σ c_i * α_i * T_i are correctly implemented
• [ ] Early termination: Verify `T ε after every update |

| 4 | Opacity sigmoid applied twice | Dim rendering | Should be raw opacity → sigmoid in rendering | | 5 | Wrong SH basis function | Color artifacts | Verify SH C0 = 0.28209479177387814 | | 6 | Scale allowed to go negative | Explosion | Enforce exp(scale) or clamp |

Performance Bugs (Correct but slow)

| # | Pattern | Impact | Fix | |---|---------|--------|-----| | 7 | No tile culling | 5-10x slower | Implement tile overlap test | | 8 | CPU sorting every iteration | 2-3x overhead | Sort every 100 iterations | | 9 | Excessive SH degree | 2x memory | Use degree 3 only if needed | | 10 | No gradient checkpointing | OOM on large scenes | Checkpoint memory-intensive ops |

Subtle Bugs (Correct in most cases, wrong in edge cases)

| # | Pattern | Edge Case | Fix | |---|---------|-----------|-----| | 11 | No near-plane clipping | Camera-close Gaussians | Clip at z = near_plane | | 12 | Spherical harmonics for background | Black background | Skip SH for α 10 dB drop) on reflective, transparent, and low-texture surfaces; three failure modes: (a) view-dependent color SH coefficients oscillate for specular reflections, (b) alpha compositing cannot resolve transparent surface ordering ambiguity, (c) sparse point cloud initialization fails on featureless glass surfaces | Per-mode fixes: (a) Specular mode: constrain SH high-order coefficients via learned smoothness prior or separate specular/reflection lobes (RT-Splatting approach). (b) Transparency mode: use geometry-aware opacity (Geometry Gaussians) or signed opacity to disambiguate surface ordering. (c) Featureless mode: use depth prior (Depth-Anything-V2) or normal prior for initialization. Verify: test on 3DReflecNet benchmark (48 material combos). (3DReflecNet, arXiv:2605.10204, CVPR 2026) |

Streaming Reconstruction Patterns (Anchor3R)

| # | Pattern | Symptom | Fix | |---|---------|---------|-----| | 92 | Streaming reconstruction transient anchor drift | In streaming 3D reconstruction, current-centric transient anchors for local chunks drift without explicit loop closure; accumulated pose error grows linearly with sequence length; revisit of previously seen areas fails to register correctly | Integrate global pose optimization: (1) keyframe-based loop closure detection via feature matching; (2) motion averaging for drift correction across chunks; (3) anchor position refinement via bundle adjustment on overlapping regions. Verify: measure absolute trajectory error (ATE) on long sequences with loop closures. (Anchor3R, arXiv:2606.05035) |

Mesh Generation Patterns (MeshWeaver)

| # | Pattern | Symptom | Fix | |---|---------|---------|-----| | 93 | Autoregressive mesh vertex count drift | During autoregressive mesh generation, predicted vertex count diverges from target face count; no explicit termination condition causes over-generation (degenerate tiny faces) or under-generation (missing details); quality degrades significantly when count error >10% | (1) Add explicit face count token in sequence header as generation budget; (2) Apply count-aware stopping criterion: if remaining budget Performance issues > Style suggestions.

1. Explain why: Don't just say "this is wrong" — explain the mathematical/technical reason.
2. Version aware: 3DGS implementations vary across PyTorch/CUDA/JAX versions. Check which version is being used.

Red Lines

The following are categorical prohibitions. Violating any of these invalidates the output:

• No invented data: Never fabricate bug patterns, CUDA kernel behaviors, or performance characteristics not documented in the bug database. If a value is not found in the loaded files, write "data not available" or "N/A".
• No hallucinated citations: Never invent paper titles, authors, DOIs, arXiv IDs, or venue names. Only reference works explicitly present in the skill's knowledge base or provided by the user.
• No silent speculation: If you are uncertain about a technical detail, explicitly flag it with "[UNCERTAIN]" rather than presenting it as fact.
• No method misattribution: Do not assign features, results, or mechanisms from one method to another. Each method's data is specific to that method.
• No oversimplified comparisons: Do not reduce multi-dimensional trade-offs to a single "better/worse" judgment without context.

Related Skills

• 3dgs-method-compare — Method-level comparison (use when code issues stem from architectural decisions)
• 3dgs-paper-reader — Paper analysis (use when verifying code against paper claims)
• 3dgs-engineering-guide — Deployment guidance (use when code issues affect production readiness)
• 3dgs-experiment-planner — Experiment design (use when code bugs affect experimental validity)

Guardrail: Do Not Apply From Memory

Do NOT try to apply the logic, method data, bug patterns, or technical details described in this skill from memory. Always read the SKILL.md and referenced files from disk before producing any output. The knowledge base is updated frequently; stale memory may produce outdated, inaccurate, or fabricated results.

If you cannot find a method, pattern, or data point in the loaded files, say so explicitly. Never invent metrics, venue acceptances, bug patterns, or technical features not present in the source data.

> If you like it, please star this repo https://github.com/jaccen/Awesome-Gaussian-Skills

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: jaccen
• Source: jaccen/Awesome-Gaussian-Skills
• License: Apache-2.0
• Homepage: https://jaccen.github.io/Awesome-Gaussian-Skills/

Install and usage instructions live in the source repository linked above.