mlsanityData to 14 Oct 2026

Ready2Blend: From Natural-Language Instructions to Composable Alignment Prompts

arXiv 2609.39365 ↗First posted 30 Sep 2026cs.CL cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Continual alignment requires LLMs to adapt to new requirements without forgetting previously acquired behaviors. Natural-language instructions are flexible and composable but offer only indirect control, whereas post-training provides stronger adaptation at the cost of repeated parameter updates. We introduce Ready2Blend, which combines the flexibility of natural language with learned alignment. AlignFormer maps each requirement to a fixed-length alignment prompt stored in a modular prompt bank, while the backbone and prior prompts remain frozen. Composability regularization transfers the semantic geometry of textual requirements into prompt space, enabling inference-time blending and reweighting. Across two practical continual alignment settings, Ready2Blend is the only frozen-backbone method that matches post-training-based alignment methods, reaching 93.1-98.5% of a joint-training reference with competitive retention, while requiring only a few prompt tokens and up to 4.3× less training time. Its modular design further enables weighted personalization and order-free composition without retraining. Code will be released upon acceptance.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 47 of its 60 references that are in the map. First posted on arXiv 30 Sep 2026.

Builds on

47 of 60 references.

References · 47

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2024The Llama 3 Herd of Modelscited by 14,715
  3. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  4. 2023BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Modelscited by 6,137
  5. 2016Overcoming catastrophic forgetting in neural networkscited by 4,787
  6. 2022Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedbackcited by 3,720
  7. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,975
  8. 2017Gradient Episodic Memory for Continual Learningcited by 1,898
  9. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,336
  10. 2022Editing Models with Task Arithmeticcited by 1,120
  11. 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,085
  12. 2024Tulu 3: Pushing Frontiers in Open Language Model Post-Trainingcited by 1,053
35 more
  1. 2026DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligencecited by 1,040
  2. 2021Learning to Prompt for Continual Learningcited by 944
  3. 2023A Comprehensive Survey of Continual Learning: Theory, Method and Applicationcited by 872
  4. 2023Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formattingcited by 817
  5. 2023The Linear Representation Hypothesis and the Geometry of Large Language Modelscited by 716
  6. 2023Safe RLHF: Safe Reinforcement Learning from Human Feedbackcited by 656
  7. 2022DualPrompt: Complementary Prompting for Rehearsal-free Continual Learningcited by 642
  8. 2026GLM-5: from Vibe Coding to Agentic Engineeringcited by 587
  9. 2023How Close is ChatGPT to Human Experts? Comparison Corpus, Evaluation, and Detectioncited by 481
  10. 2022CODA-Prompt: COntinual Decomposed Attention-based Prompting for Rehearsal-Free Continual Learningcited by 450
  11. 2023Learning to Compress Prompts with Gist Tokenscited by 390
  12. 2021Rainbow Memory: Continual Learning with a Memory of Diverse Samplescited by 273
  13. 2023The Unlocking Spell on Base LLMs: Rethinking Alignment via In-Context Learningcited by 257
  14. 2023Rewarded soups: towards Pareto-optimal alignment by interpolating weights fine-tuned on diverse rewardscited by 243
  15. 2023Progressive Prompts: Continual Learning for Language Modelscited by 197
  16. 2024Aligning with Human Judgement: The Role of Pairwise Preference in Large Language Model Evaluatorscited by 149
  17. 2024Arithmetic Control of LLMs for Diverse User Preferences: Directional Preference Alignment with Multi-Objective Rewardscited by 148
  18. 2024FineSurE: Fine-grained Summarization Evaluation using LLMscited by 83
  19. 2024Towards Lifelong Learning of Large Language Models: A Surveycited by 64
  20. 2024DeAL: Decoding-time Alignment for Large Language Modelscited by 55
  21. 2024Consistent Prompting for Rehearsal-Free Continual Learningcited by 54
  22. 2024Learning to Summarize from LLM-generated Feedbackcited by 24
  23. 2024HFT: Half Fine-Tuning for Large Language Modelscited by 18
  24. 2025RainbowPrompt: Diversity-Enhanced Prompt-Evolving for Continual Learningcited by 10
  25. 2024Q-Tuning: Queue-based Prompt Tuning for Lifelong Few-shot Language Learningcited by 9
  26. 2023Enhancing Abstractiveness of Summarization Models through Calibrated Distillationcited by 8
  27. 2023One Size Fits All for Semantic Shifts: Adaptive Prompt Tuning for Continual Learningcited by 8
  28. 2025LifeAlign: Lifelong Alignment for Large Language Models with Memory-Augmented Focalized Preference Optimizationcited by 7
  29. 2025Adaptive Helpfulness-Harmlessness Alignment with Preference Vectorscited by 6
  30. 2025Sequence Transferability and Task Order Selection in Continual Learningcited by 5
  31. 2025Tuning-Free Personalized Alignment via Trial-Error-Explain In-Context Learningcited by 4
  32. 2025Mitigating Catastrophic Forgetting in Target Language Adaptation of LLMs via Source-Shielded Updatescited by 4
  33. 2026Mitigating Task-Order Sensitivity and Forgetting via Hierarchical Second-Order Consolidationcited by 1
  34. 2026Alignment Tuning for Large Language Models: A Data-Centric Lens on Alignment Data Pipelinescited by 1
  35. 2026STAGE: Controlled Objective Admission for Multi-Preference LLM Alignmentcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026ECLIPTICA -- A Framework for Switchable LLM Alignment via CITA - Contrastive Instruction-Tuned Alignment
  2. 2024Dynamic Rewarding with Prompt Optimization Enables Tuning-free Self-Alignment of Language Models
  3. 2026AlignTune: Modular Toolkit for Post-Training Alignment of Large Language Models
  4. 2026CAP: Controllable Alignment Prompting for Unlearning in LLMs
  5. 2024Is Free Self-Alignment Possible?
  6. 2023Aligner: One Global Token is Worth Millions of Parameters When Aligning Large Language Models
  7. 2024Aligner: Efficient Alignment by Learning to Correct
  8. 2025On Almost Surely Safe Alignment of Large Language Models at Inference-Time