Papers › Resolving Task Confusion in Dynamic Expansion Architectures for Class Incremental Learning
Resolving Task Confusion in Dynamic Expansion Architectures for Class Incremental Learning
Bingchen Huang, Zhineng Chen, Peng Zhou, Jiayin Chen, Zuxuan Wu
The dynamic expansion architecture is becoming popular in class incremental learning, mainly due to its advantages in alleviating catastrophic forgetting. However, task confusion is not well assessed within this framework, e.g., the discrepancy between classes of different tasks is not well learned (i.e., inter-task confusion, ITC), and certain priority is still given to the latest class batch (i.e., old-new confusion, ONC). We empirically validate the side effects of the two types of confusion. Meanwhile, a novel solution called Task Correlated Incremental Learning (TCIL) is proposed to encourage discriminative and fair feature utilization across tasks. TCIL performs a multi-level knowledge distillation to propagate knowledge learned from old tasks to the new one. It establishes information flow paths at both feature and logit levels, enabling the learning to be aware of old classes. Besides, attention mechanism and classifier re-scoring are applied to generate more fair classification scores. We conduct extensive experiments on CIFAR100 and ImageNet100 datasets. The results demonstrate that TCIL consistently achieves state-of-the-art accuracy. It mitigates both ITC and ONC, while showing advantages in battle with catastrophic forgetting even no rehearsal memory is reserved.
In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.
Code
Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.
Code Syntology ran Syntology
Not run by Syntology. Nothing on this page verifies that the listed code works.
Tasks
1 archive task tag without a task page not shown.
Results from the paper archive 2025-07-28
| Task | Dataset | Model | Metric | Value | Rank at snapshot | Leaderboard | Report |
|---|---|---|---|---|---|---|---|
| Incremental Learning | CIFAR-100 - 50 classes + 10 steps of 5 classes | TCIL | Average Incremental Accuracy | 73.72 | #1 of 13 | Archive leaderboard | report |
| Incremental Learning | CIFAR-100 - 50 classes + 10 steps of 5 classes | TCIL-Lite | Average Incremental Accuracy | 73.50 | #2 of 13 | Archive leaderboard | report |
| Incremental Learning | CIFAR-100 - 50 classes + 2 steps of 25 classes | TCIL | Average Incremental Accuracy | 76.42 | #1 of 4 | Archive leaderboard | report |
| Incremental Learning | CIFAR-100 - 50 classes + 2 steps of 25 classes | TCIL-Lite | Average Incremental Accuracy | 74.95 | #2 of 4 | Archive leaderboard | report |
| Incremental Learning | CIFAR-100 - 50 classes + 5 steps of 10 classes | TCIL | Average Incremental Accuracy | 74.88 | #1 of 15 | Archive leaderboard | report |
| Incremental Learning | CIFAR-100 - 50 classes + 5 steps of 10 classes | TCIL-Lite | Average Incremental Accuracy | 74.30 | #2 of 15 | Archive leaderboard | report |
| Incremental Learning | CIFAR-100-B0(5steps of 20 classes) | TCIL | Average Incremental Accuracy | 77.72 | #2 of 10 | Archive leaderboard | report |
| Incremental Learning | CIFAR-100-B0(5steps of 20 classes) | TCIL-Lite | Average Incremental Accuracy | 76.96 | #3 of 10 | Archive leaderboard | report |
| Incremental Learning | CIFAR100-B0(10steps of 10 classes) | TCIL | Average Incremental Accuracy | 77.30 | #2 of 6 | Archive leaderboard | report |
| Incremental Learning | CIFAR100-B0(10steps of 10 classes) | TCIL-Lite | Average Incremental Accuracy | 76.74 | #3 of 6 | Archive leaderboard | report |
| Incremental Learning | CIFAR100B020Step(5ClassesPerStep) | TCIL-Lite | Average Incremental Accuracy | 75.47 | #2 of 5 | Archive leaderboard | report |
| Incremental Learning | CIFAR100B020Step(5ClassesPerStep) | TCIL | Average Incremental Accuracy | 75.11 | #3 of 5 | Archive leaderboard | report |
| Incremental Learning | ImageNet100 - 10 steps | TCIL | # M Params | 116.54 | #4 of 13 | Archive leaderboard | report |
| Incremental Learning | ImageNet100 - 10 steps | TCIL | Average Incremental Accuracy | 77.66 | #4 of 13 | Archive leaderboard | report |
| Incremental Learning | ImageNet100 - 10 steps | TCIL | Average Incremental Accuracy Top-5 | 94.17 | #4 of 13 | Archive leaderboard | report |
| Incremental Learning | ImageNet100 - 10 steps | TCIL | Final Accuracy | 67.34 | #4 of 13 | Archive leaderboard | report |
| Incremental Learning | ImageNet100 - 10 steps | TCIL | Final Accuracy Top-5 | 88.84 | #4 of 13 | Archive leaderboard | report |
| Incremental Learning | ImageNet100 - 10 steps | TCIL-Lite | # M Params | 26.36 | #5 of 13 | Archive leaderboard | report |
| Incremental Learning | ImageNet100 - 10 steps | TCIL-Lite | Average Incremental Accuracy | 77.50 | #5 of 13 | Archive leaderboard | report |
| Incremental Learning | ImageNet100 - 10 steps | TCIL-Lite | Average Incremental Accuracy Top-5 | 93.60 | #5 of 13 | Archive leaderboard | report |
| Incremental Learning | ImageNet100 - 10 steps | TCIL-Lite | Final Accuracy | 67.30 | #5 of 13 | Archive leaderboard | report |
| Incremental Learning | ImageNet100 - 10 steps | TCIL-Lite | Final Accuracy Top-5 | 87.94 | #5 of 13 | Archive leaderboard | report |
Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.
Methods
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections