Research map / Continual learning / Continual fine-tuning of LLMs
Parameterizing Context: Unleashing the Power of Parameter-Efficient Fine-Tuning and In-Context Tuning for Continual Table Semantic Parsing
Yongrui Chen, Shenyu Zhang, Guilin Qi, Xinnan Guo
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| add teacher (ICT) with student distillation to prompt-tuning (ablation) | PEFT (T5-base) · Spider-Stream | TA: 65.7 -> 67.7 | Improved |
| add teacher (ICT) with student distillation to prompt-tuning (ablation) | PEFT (T5-base) · Combined-Stream | TA: 63.8 -> 66.4 | Improved |
| add student prompt-tuning (compress teacher into prompts) (ablation) | C3 · Combined-Stream | TA: 59.6 -> 66.4 | Improved |
| add student prompt-tuning (compress teacher into prompts) (ablation) | C3 · Spider-Stream | TA: 64.9 -> 67.7 | Improved |
| use in-context tuning for the teacher instead of fine-tuning (ablation) | C3 · Spider-Stream | TA: 65.6 -> 67.7 | Improved |
| use in-context tuning for the teacher instead of fine-tuning (ablation) | C3 · Combined-Stream | TA: 65.3 -> 66.4 | Improved |
| add initial task adaptation (fine-tune backbone on first task) (ablation) | C3 · Spider-Stream | TA: 19.4 -> 67.7 | Improved |
| add initial task adaptation (fine-tune backbone on first task) (ablation) | C3 · Combined-Stream | TA: 35.8 -> 66.4 | Improved |
| initialize prompts from previous task prompt instead of shared P* (ablation) | C3 · Spider-Stream | TA: 67.7 -> 67.0 | Hurt |
| initialize prompts from previous task prompt instead of shared P* (ablation) | C3 · Combined-Stream | TA: 66.4 -> 66.0 | Hurt |
| replace T5-large teacher with text-davinci-003 (no tuning) | C3 (T5-large student) · Spider-Stream | TA: 71.1 -> 71.3 | No clear effect |
| replace T5-large teacher with text-davinci-003 (no tuning) | C3 (T5-large student) · Combined-Stream | TA: 68.3 -> 67.6 | Hurt |
Closest papers by meaning
- Filling Memory Gaps: Enhancing Continual Semantic Parsing via SQL Syntax Variance-Guided LLMs Without Real Data Replay (AAAI 2025 · similarity 0.92)
- Learn more, but bother less: parameter efficient continual learning (NeurIPS 2024 · similarity 0.87)
- Adaptive Prompting for Continual Relation Extraction: A Within-Task Variance Perspective (AAAI 2025 · similarity 0.86)
- Scalable Language Model with Generalized Continual Learning (ICLR 2024 · similarity 0.86)
- Train-Attention: Meta-Learning Where to Focus in Continual Knowledge Learning (NeurIPS 2024 · similarity 0.86)
- NeedleInATable: Exploring Long-Context Capability of Large Language Models towards Long-Structured Tables (NeurIPS 2025 · 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.
