The day in conversation for 2026-11-05
discussion
Transcript
The discussion: Rosa: The 60% parameter reduction from Deep SSM Compression sounds impressive for the IMDb task, but I’m curious how that compression holds up when we move beyond shallow layers and into deeper architectures; does the approximation error accumulate faster than we expect? Dev: That's exactly what worries me. My concern with those gradient-based MOR methods optimizing the compression upper bound is that they might find a local optimum in the search space, meaning we get a compressed model that looks good on IMDb but fails spectacularly on tasks where feature interaction is more complex, like sentiment analysis requiring nuanced context. Taro: I read about Symphony and the wave energy systems—that ninety percent absorption rate implies incredible efficiency for non-optimization control, which feels fundamentally different from the parameter reduction we see in Deep SSM. What’s the trade-off on that power absorption metric when you consider the physical constraints of the specific wave frequency ranges being tested? Rosa: The constraint is often overlooked in those initial evaluations. Symphony achieved ninety percent absorption under idealized conditions, but how does that performance shift if the water viscosity changes by just five percent, or if there’s significant non-linear damping introduced into the circuit model? Dev: That connects back to my work on battery prediction, which uses a hybrid neural correction layered onto a Thevenin circuit. We’re dealing with dynamic environments where those physical parameters aren't fixed constants; the robustness of that
More episodes
- 2608.13438-ContactGuard: Pre-Contact Execution Monitoring with Action-Conditioned Latent World Models
- 2610.34593-Future Prediction as Training-Only Video Supervision for Pretrained VLA Policies
- 2610.35385-EmbodiedLab: Beyond Manipulation, Benchmarking Embodied Agents That Act to Know
- 2610.36378-Data-Driven Hamiltonian for Neural Optimal Control
- 2610.37275-A Heuristic for Scheduling Autonomous Vehicles at Intersections
- 2610.37288-Dynamic Coordination of Autonomous Vehicles at Intersections
- 2610.37632-Real-Time Coordination of Autonomous Vehicles at Intersections
- 2611.13779-Softness is something you also need! Why compliance helps control: information metrics of a passive walker with compliant ankle
- 2611.13411-Residual-Corrected Physics-Informed Neural Network KKL Observers with Input-to-State Stability Guarantees
- 2611.14488-Is Your Robot's Memory Just a Clock? Temporal Shortcuts in Memory-Augmented VLA Policies