# Joint Class-Time Learning for Video Classification with Multi-Instance Partial-Label Learning

arXiv: https://arxiv.org/abs/2610.06234 · HTML: https://mlsanity.com/paper/2610.06234/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 20 of its 31 references that are in the map.
- First posted on arXiv 5 Oct 2026.

## Builds on (most cited first, 20 of 20)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [SlowFast Networks for Video Recognition](https://mlsanity.com/paper/1812.03982/) (2018)
- [Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality](https://mlsanity.com/paper/2405.21060/) (2024)
- [VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking](https://mlsanity.com/paper/2303.16727/) (2023)
- [UntrimmedNets for Weakly Supervised Action Recognition and Detection](https://mlsanity.com/paper/1703.03329/) (2017)
- [FineAction: A Fine-Grained Video Dataset for Temporal Action Localization](https://mlsanity.com/paper/2105.11107/) (2021)
- [ASM-Loc: Action-aware Segment Modeling for Weakly-Supervised Temporal Action Localization](https://mlsanity.com/paper/2203.15187/) (2022)
- [SoLar: Sinkhorn Label Refinery for Imbalanced Partial-Label Learning](https://mlsanity.com/paper/2209.10365/) (2022)
- [Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature](https://mlsanity.com/paper/2303.12332/) (2023)
- [Proposal-Based Multiple Instance Learning for Weakly-Supervised Temporal Action Localization](https://mlsanity.com/paper/2305.17861/) (2023)
- [Candidate Pseudolabel Learning: Enhancing Vision-Language Models by Prompt Tuning with Unlabeled Data](https://mlsanity.com/paper/2406.10502/) (2024)
- [Realistic Evaluation of Deep Partial-Label Learning Algorithms](https://mlsanity.com/paper/2502.10184/) (2025)
- [Disambiguated Attention Embedding for Multi-Instance Partial-Label Learning](https://mlsanity.com/paper/2305.16912/) (2023)
- [Weakly Supervised Temporal Action Localization via Dual-Prior Collaborative Learning Guided by Multimodal Large Language Models](https://mlsanity.com/paper/2411.08466/) (2024)
- [Multi-Instance Partial-Label Learning: Towards Exploiting Dual Inexact Supervision](https://mlsanity.com/paper/2212.08997/) (2022)
- [Rethinking Pseudo-Label Guided Learning for Weakly Supervised Temporal Action Localization from the Perspective of Noise Correction](https://mlsanity.com/paper/2501.11124/) (2025)
- [Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer](https://mlsanity.com/paper/2504.14860/) (2025)
- [Multi-Instance Partial-Label Learning with Margin Adjustment](https://mlsanity.com/paper/2501.12597/) (2025)
- [Exploiting Conjugate Label Information for Multi-Instance Partial-Label Learning](https://mlsanity.com/paper/2408.14369/) (2024)
- [Calibratable Disambiguation Loss for Multi-Instance Partial-Label Learning](https://mlsanity.com/paper/2512.17788/) (2025)

## Abstract (arXiv)

Multi-instance partial-label learning (MIPL) addresses inexact supervision in both the instance and label spaces, which can be applied to video classification. However, bag-level labels do not explicitly supervise the correspondence between candidate classes and temporal evidence. We propose ours, which couples label disambiguation with temporal evidence allocation through a joint class--time assignment. Occupancy-regularized spherical matching associates contextualized video features while learning nonuniform temporal mass and discouraging excessive concentration. During training, candidate-restricted inference recomputes the assignment within the candidate label set. A dual-marginal KL projection then constructs a structured teacher that incorporates momentum-refined class beliefs while preserving the proposal's temporal occupancy. A single plan-level KL objective aligns the full-space predictor with this teacher. Our analysis characterizes when candidate re-solving differs from masking and shows that, under the stated construction, the joint objective decomposes into class-marginal and class-conditional temporal supervision. We construct VCMIPL benchmarks from Breakfast, DoTA, and FineAction using model-generated candidate labels and evaluate the method across four feature representations. Extensive experimental results demonstrate that PIVOTMIPL outperforms existing MIPL algorithms in both effectiveness and efficiency.