UPSTREAM PR #18406: mtmd: clarify that we no longer accept AI-generated PRs

[loci-dev]

Mirrored from ggml-org/llama.cpp#18406

I demand other maintainers to either accept this or ggml-org/llama.cpp#18388

AI-generated PRs in mtmd are often, if not always, lead to sub-optimal, low-quality code. These PRs are usually huge, take me too much time to review / optimize, adding more mental strain and make the productivity goes negative.

Therefore, I would like to explicitly reject PRs that are fully or predominantly AI-generated. Using AI as an assistive tool is always welcomed.

Important

This message does not prevent contributors to push such PRs in the future, but serves as an explicit statement in case someone asks

Some examples:

• add model: Qwen3-Omni-30B-A3B-Instruct ggml-org/llama.cpp#18404
• Add Gemma3n multimodal support with MobileNetV5 vision encoder ggml-org/llama.cpp#18256
• mtmd: add GLM4V multimodal model with conversion support ggml-org/llama.cpp#17998
• mtmd: add Eagle2-VL vision and projector support ggml-org/llama.cpp#17360

[loci-review]

Explore the complete analysis inside the Version Insights

Perfect! I was able to retrieve the summary report for your project. Here's what the analysis shows:

Summary Report for llama.cpp PR #718

Project Information:

• Repository: llama.cpp (owner: auroralabs-loci)
• Pull Request: UPSTREAM PR #18406: mtmd: clarify that we no longer accept AI-generated PRs #718
• Comparing versions:
  • Base: 16a6f09b-dc7d-4572-a33f-2adf1a63a275
  • Target: 5a6eab03-544a-42fb-9635-19e5a70e5fcc

Key Findings:

✅ No significant performance changes detected - The analysis found no modified functions with performance changes greater than 2% in either:

• Response Time
• Throughput Time

Conclusion:

This PR appears to be performance-neutral, meaning the code changes don't introduce any performance regressions or improvements beyond the 2% threshold. This is a positive result indicating that the changes maintain stable performance characteristics of the application.