aakanksha

Finally, we find that model coefficients while merging directly impact the performance on key metrics — i.e., higher safety weighting leads to better alignment but at the cost of general performance after a certain point 📈📉 https://t.co/PfnzvfP13J

Our work is an important step in understanding the merits of model merging in a multi-task setting while optimizing for multiple objectives! 🎯🉑📚 We hope this spurs additional collaborations. 📜 https://t.co/4iptoe905Z