
Fin ML: How to Align a Model—and How to Break It
Düzenleyen: Harry McEnaney
15 Temmuz 2026 Çarşamba
18:30 GMT+1
15 Temmuz 2026 Çarşamba
20:30 GMT+1
Ücret
Ücretsiz
Katılım ücretsiz
Side Events
1Fin ML: How to Align a Model—and How to Break It boyunca London olan tüm etkinlikler: şehirdeyken uğrayabileceğin yan buluşmalar, gün gün.
Etkinlik Hakkında
Fin builds AI-native customer products. Join two of Fin's ML scientists at our Old Street London office on July 15, 2026, for original talks on post-training alignment and inference-time misalignment — followed by food, drinks, and conversation. Doors open at 6:00pm, talks start at 6:30pm. 6:00pm - 6:30pm: Arrival 6:30pm - 7:00pm: Vivek Anand - Senior Machine Learning Scientist Building Aligned Models: The Landscape of Post-Training A pre-trained model knows almost everything and is aligned to almost nothing. This talk explores how we turn raw capability into a model that we can trust, shaped first to follow instructions, then tuned to value what we value. We'll walk through the post-training landscape that makes this alignment possible. 7:00pm - 7:30pm: Gilad Tsur - Senior Machine Learning Scientist From Aligned Models to Misaligned Behaviours: Examining misalignment at inference time We explore several attempts to get models to produce behaviours that they were trained not to produce. We discuss what worked, what didn't work, and what we can learn from this. 7:30pm - 8:30pm: Food, drinks and networking 8:30pm: Finish These talks draw on work we're doing in production at Fin — we'd love to discuss it with people thinking about the same problems.
This is a limited-capacity session.
Please submit your interest and someone from the team will confirm your attendance.
Mekan Bilgileri
Intercom
Intercom, London, England
Çevreyi keşfet
Ziyaretçilere Ücretsiz
Düzenleyen
Önce yukarıdan "Katılacağım"a bas. İşaretleyince bu etkinliğe katılan sana uygun kişilerle tanışabilirsin.