Which Mutual-Information Representation Learning Objectivesare Sufficientfor Control?
–Neural Information Processing Systems
Figure 4: (left) Originalcatchergame: agent (grey paddle) movesleftorrighttocatchfruit (yellow square) thatfalls. Ouraim representations, objectiv 8 Err.
Neural Information Processing Systems
Feb-11-2026, 12:15:47 GMT