Evolution Beats Backprop for LLM Fine-Tuning
Researchers demonstrate that evolutionary algorithms can outperform traditional backpropagation methods for fine-tuning large language models, offering a
Someone found that evolutionary strategies can replace backpropagation for fine-tuning language models, which sounds ridiculous but actually works.
The paper at https://arxiv.org/abs/2509.24372 showed that 30 random gaussian perturbations can approximate gradients well enough to beat GRPO on RLVR tasks. Zero overfitting, way faster training since there’s no backward pass.
A developer tested it themselves and got it working: https://github.com/Green0-0/propagate
The repo now includes LoRA support and pass@k training. Pretty wild that you can fine-tune models by just adding random noise and seeing what works better - no gradient calculations needed.
Worth checking out if standard fine-tuning feels too slow or memory-heavy. The approach trades computational precision for speed, and apparently the tradeoff works surprisingly well for RL tasks.
Related Tips
GPU Shortage Tracker Shows Grim Hardware Upgrade Outlook
The GPU Shortage Tracker reveals a bleak outlook for hardware upgrades as graphics card availability remains severely limited and prices continue to climb well
Snapchat Scammers Use Open-Source LLMs for Sextortion
Scammers are exploiting open-source large language models on Snapchat to automate sextortion schemes, targeting vulnerable users through AI-generated
Samsung's SOCAMM2: Replaceable LPDDR5X for AI Servers
Samsung unveils SOCAMM2, a revolutionary replaceable LPDDR5X memory module designed specifically for AI servers, enabling easier upgrades and improved