Cambridge Machine Learning Group
@cambridgemlg.bsky.social
We’re so back.
You will be a great fit if you have expertise in technical methods for human-AI cooperation, reinforcement learning, or causal inference.
Deadline: 21 January 2025
Deadline: 21 January 2025
December 18, 2024 at 3:02 PM
You will be a great fit if you have expertise in technical methods for human-AI cooperation, reinforcement learning, or causal inference.
Deadline: 21 January 2025
Deadline: 21 January 2025
Come meet us at the posters if you want to chat about any of these papers!
December 11, 2024 at 6:41 PM
Come meet us at the posters if you want to chat about any of these papers!
And lastly, James Requeima & John Bronskill will be presenting:
“LLM Processes: Numerical Predictive Distributions Conditioned on Natural Language”
at 16:30 — Poster Session 2 East, #2901
arxiv.org/abs/2405.128...
“LLM Processes: Numerical Predictive Distributions Conditioned on Natural Language”
at 16:30 — Poster Session 2 East, #2901
arxiv.org/abs/2405.128...
LLM Processes: Numerical Predictive Distributions Conditioned on Natural Language
Machine learning practitioners often face significant challenges in formally integrating their prior knowledge and beliefs into predictive models, limiting the potential for nuanced and context-aware ...
arxiv.org
December 11, 2024 at 6:40 PM
And lastly, James Requeima & John Bronskill will be presenting:
“LLM Processes: Numerical Predictive Distributions Conditioned on Natural Language”
at 16:30 — Poster Session 2 East, #2901
arxiv.org/abs/2405.128...
“LLM Processes: Numerical Predictive Distributions Conditioned on Natural Language”
at 16:30 — Poster Session 2 East, #2901
arxiv.org/abs/2405.128...
Dima Krasheninnikov will be presenting:
“Stress-Testing Capability Elicitation with Password-Locked Models”
See it today at 11am in East Exhibit Hall A, #2403
arxiv.org/abs/2405.19550
“Stress-Testing Capability Elicitation with Password-Locked Models”
See it today at 11am in East Exhibit Hall A, #2403
arxiv.org/abs/2405.19550
Stress-Testing Capability Elicitation With Password-Locked Models
To determine the safety of large language models (LLMs), AI developers must be able to assess their dangerous capabilities. But simple prompting strategies often fail to elicit an LLM's full capabilit...
arxiv.org
December 11, 2024 at 6:40 PM
Dima Krasheninnikov will be presenting:
“Stress-Testing Capability Elicitation with Password-Locked Models”
See it today at 11am in East Exhibit Hall A, #2403
arxiv.org/abs/2405.19550
“Stress-Testing Capability Elicitation with Password-Locked Models”
See it today at 11am in East Exhibit Hall A, #2403
arxiv.org/abs/2405.19550
@mashtro.bsky.social and Cristiana Diaconu will present:
“Approximately Equivariant Neural Processes”
at 16:30 -- Poster Session 2 East, #4903
arxiv.org/abs/2406.13488
“Approximately Equivariant Neural Processes”
at 16:30 -- Poster Session 2 East, #4903
arxiv.org/abs/2406.13488
Approximately Equivariant Neural Processes
Equivariant deep learning architectures exploit symmetries in learning problems to improve the sample efficiency of neural-network-based models and their ability to generalise. However, when modelling...
arxiv.org
December 11, 2024 at 6:40 PM
@mashtro.bsky.social and Cristiana Diaconu will present:
“Approximately Equivariant Neural Processes”
at 16:30 -- Poster Session 2 East, #4903
arxiv.org/abs/2406.13488
“Approximately Equivariant Neural Processes”
at 16:30 -- Poster Session 2 East, #4903
arxiv.org/abs/2406.13488