Alice 85jj

Possible Contexts:

5. Experiments

5.1 Benchmarks

| Dataset | # Tasks | Classes / Task | Input Size | |-------------|------------|-------------------|----------------| | Split‑CIFAR‑100 | 10 | 10 | 32 × 32 | | CORe50 (NC) | 9 | 5‑10 | 128 × 128 | | TinyImageNet‑Continual | 20 | 20 | 64 × 64 | | Robo‑Manip (Lifelong) | 7 | 6 (objects) | 224 × 224 + proprioception | alice 85jj

Results:

1. Adaptive Lateral Inhibition (ALICE) – computes a task‑relevance mask M that attenuates channels irrelevant to the current context.
2. Semantic Junction (S‑Junction) – aggregates F across spatial dimensions, producing a compact semantic embedding z_s.
3. Contextual Junction (C‑Junction) – encodes the task descriptor τ (e.g., language prompt, sensor metadata) via a shallow Transformer, yielding z_c.

Additionally, what kind of article are you looking for? Would you like it to be: Possible Contexts : 5