Research map / Continual learning / Continual fine-tuning of LLMs
Enhancing Pretrained Model-based Continual Representation Learning via Guided Random Projection
Ruilin Li, Heming Zou, Xiufeng Yan, Zheming Liang, Jie Yang, Chenliang Li, Xue Yang
This paper tried a common change: Larger batch size
increase batch size s tenfold · CIL benchmarks · A_last No clear effect
What 785 accepted papers report for this change
- 469 improved
- 107 no clear effect
- 48 mixed
- 161 hurt
- Predicting the Emergence of Induction Heads in Language Model Pretraining ICML 2026 · increase batch size (fixed context 1024) · Natural data (CC100) · induction head emergence point (updates): rho=-1.0, p<0.001Improved
- When RL Meets Adaptive Speculative Training: A Unified Training-Serving System ICML 2026 · Increase batch size from 1 to 32 · MiniMax M2.1 FP8 · speedup: 1.57x -> 1.25xHurt
- Untied Ulysses: Memory-Efficient Context Parallelism via Headwise Chunking ICML 2026 · increase per-device batch size to 10 using freed memory · Llama 3-8B, 128K · throughput (TPS): 2464.10 (Ulysses) -> 2526.77 (UPipe)Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace random RPL with MGSM data-guided RPL construction | RPL-based CIL · ImageNet-R (B0 Inc-10) | A_last: 75.07 -> 77.12 (vs best baseline APT) | Improved |
| replace SCSM greedy criterion with MGSM criterion at 10,000 hidden units (ablation) | SCL-MGSM · four CIL benchmarks | A_last: approximately 5%-10% gain over SCSM | Improved |
| replace random initialization with MGSM at 10,000 hidden units, ImageNet-A (ablation) | SCL-MGSM · ImageNet-A | A_last | Improved |
| SCSM vs random initialization (ablation) | RPL construction · frozen PTM features | A_last | Hurt |
| use adaptive xi instead of fixed xi (0.08 or 0.008) (ablation) | SCL-MGSM · four CIL benchmarks | A_last | Improved |
| increase batch size s tenfold (ablation) | SCL-MGSM · CIL benchmarks | A_last | No clear effect |
| enable first-session adaptation (FSA) (ablation) | RPL-based methods · CIL benchmarks | gain over joint linear probe | Improved |
| swap ViT-B/16-IN21K for DINO-v2 | SCL-MGSM · ImageNet-R (B0 Inc-10) | A_last: 77.12 -> 86.40 | Improved |
| MGSM RPL on ResNet-101 backbone | SCL-MGSM · ImageNet-A (B0 Inc-10) | A_last: 20.08 vs LoRanPAC 21.94 | Hurt |
| MGSM RPL with moderate lambda instead of large lambda for RI (ablation) | SCL-MGSM · ImageNet-A | condition number / ||P_t||_F | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- AnaCP — AnaCP: Toward Upper-Bound Continual Learning via Analytic Contrastive Projection (NeurIPS 2025) · better on 10 of 10 shared results
- LoRanPAC — LoRanPAC: Low-rank Random Features and Pre-trained Models for Bridging Theory and Practice in Continual Learning (ICLR 2025) · better on 6 of 6 shared results
- RanPAC — RanPAC: Random Projections and Pre-trained Models for Continual Learning (NeurIPS 2023) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Self-Expansion of Pre-trained Models with Mixture of Adapters for Continual Learning (CVPR 2025 · similarity 0.87)
- RanDumb: Random Representations Outperform Online Continually Learned Representations (NeurIPS 2024 · similarity 0.87)
- Non-exemplar Online Class-Incremental Continual Learning via Dual-Prototype Self-Augment and Refinement (AAAI 2024 · similarity 0.86)
- Enhancing Online Continual Learning with Plug-and-Play State Space Model and Class-Conditional Mixture of Discretization (CVPR 2025 · similarity 0.86)
- Enhancing Visual Continual Learning with Language-Guided Supervision (CVPR 2024 · similarity 0.86)
- Model Sensitivity Aware Continual Learning (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
