community
active
leiden_hybrid_concepts
label: haiku
community:leiden_hybrid_concepts-run4-c14-c4Multi-task learning gradient balancing
Techniques for combining loss-scale and gradient-magnitude weighting to improve multi-task dense prediction on NYUv2 benchmark.
3 members. Each node is clickable.
Loading graph…
Drawn from 1 source
The papers/notes whose extracted claims & findings make up this cluster.
Bridges (2)
Other communities that share members with this one — cross-cutting threads or papers that sit at the seam between two themes.
Findings (2)
- Combining loss-scale and gradient-magnitude balancing achieves Δp = +1.15±0.16 on NYUv2.Full DB-MTL ablation result.
- On NYUv2, EW suffers a drop in surface normal prediction (mean angle error 23.57 vs STL 21.99, within 11.25° 35.04 vs 39.04).Task balancing issue where surface normal prediction degrades under EW.