mlsanityData to 14 Oct 2026

CantTalkAboutThis: Aligning Language Models to Stay on Topic in Dialogues

arXiv 2404.03820 ↗First posted 4 Apr 2024cs.CLcited by 15

On the map: LLM reasoning → Reasoning benchmarks and evaluation

also in 1 more
Abstract (arXiv)

Recent advancements in instruction-tuning datasets have predominantly focused on specific tasks like mathematical or logical reasoning. There has been a notable gap in data designed for aligning language models to maintain topic relevance in conversations - a critical aspect for deploying chatbots to production. We introduce the CantTalkAboutThis dataset to help language models remain focused on the subject at hand during task-oriented interactions. It consists of synthetic dialogues on a wide range of conversation topics from different domains. These dialogues are interspersed with distractor turns that intentionally divert the chatbot from the predefined topic. Fine-tuning language models on this dataset helps make them resilient to deviating from the role assigned and improves their ability to maintain topical coherence compared to general-purpose instruction-tuned LLMs like GPT-4-turbo and Mixtral-Instruct. Additionally, preliminary observations suggest that training models on this dataset also enhance their performance on fine-grained instruction following tasks, including safety alignment.

Key facts. Cited by 15 papers in the map (data through 14 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 32 of its 54 references that are in the map. First posted on arXiv 4 Apr 2024.

Builds on

32 of 54 references.

References · 32

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  3. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,170
  4. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  5. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  6. 2020Scaling Laws for Neural Language Modelscited by 6,586
  7. 2022Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedbackcited by 3,720
  8. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,724
  9. 2024Mixtral of Expertscited by 1,823
  10. 2020GLU Variants Improve Transformercited by 1,804
  11. 2023Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversationscited by 1,207
  12. 2023LIMA: Less Is More for Alignmentcited by 1,100
20 more
  1. 2023BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Datasetcited by 860
  2. 2018MultiWOZ -- A Large-Scale Multi-Domain Wizard-of-Oz Dataset for Task-Oriented Dialogue Modellingcited by 766
  3. 2023The Flan Collection: Designing Data and Methods for Effective Instruction Tuningcited by 739
  4. 2023OpenAssistant Conversations -- Democratizing Large Language Model Alignmentcited by 721
  5. 2023Textbooks Are All You Need II: phi-1.5 technical reportcited by 565
  6. 2022Unnatural Instructions: Tuning Language Models with (Almost) No Human Laborcited by 397
  7. 2023NeMo Guardrails: A Toolkit for Controllable and Safe LLM Applications with Programmable Railscited by 365
  8. 2023ToxicChat: Unveiling Hidden Challenges of Toxicity Detection in Real-World User-AI Conversationcited by 231
  9. 2017A Survey on Dialogue Systems: Recent Advances and New Frontierscited by 162
  10. 2023HelpSteer: Multi-attribute Helpfulness Dataset for SteerLMcited by 142
  11. 2023SimpleSafetyTests: a Test Suite for Identifying Critical Safety Risks in Large Language Modelscited by 76
  12. 2019Out-of-domain Detection for Natural Language Understanding in Dialog Systemscited by 62
  13. 2022Dialog Inpainting: Turning Documents into Dialogscited by 56
  14. 2023Can LLMs Follow Simple Rules?cited by 46
  15. 2024Book2Dial: Generating Teacher-Student Interactions from Textbooks for Cost-Effective Development of Educational Chatbotscited by 25
  16. 2021ROPE: Reading Order Equivariant Positional Encoding for Graph-based Document Information Extractioncited by 22
  17. 2024SoFA: Shielded On-the-fly Alignment via Priority Rule Followingcited by 18
  18. 2024Suppressing Pink Elephants with Direct Principle Feedbackcited by 15
  19. 2023Dialogizer: Context-aware Conversational-QA Dataset Generation from Textual Sourcescited by 9
  20. 2023CONSCENDI: A Contrastive and Scenario-Guided Distillation Approach to Guardrail Models for Virtual Assistantscited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 1; 2025: 7; 2026: 7.

15 papers.

Sort:

By area:

  1. 2024Nemotron-4 340B Technical Reportcited by 144
  2. 2025Aegis2.0: A Diverse AI Safety Dataset and Risks Taxonomy for Alignment of LLM Guardrailscited by 132
  3. 2025DynaGuard: A Dynamic Guardian Model With User-Defined Policiescited by 19
  4. 2025Understanding Why ChatGPT Outperforms Humans in Visualization Design Advicecited by 5
  5. 2026MENASpeechBank: A Reference Voice Bank with Persona-Conditioned Multi-Turn Conversations for AudioLLMscited by 3
  6. 2025DeepThink: Aligning Language Models with Domain-Specific User Intentscited by 2
  7. 2025One-Pass to Reason: Token Duplication and Block-Sparse Mask for Efficient Fine-Tuning on Multi-Turn Reasoningcited by 1
  8. 2025EnSToM: Enhancing Dialogue Systems with Entropy-Scaled Steering Vectors for Topic Maintenancecited by 1
  9. 2025CogGen: A Learner-Centered Generative AI Architecture for Intelligent Tutoring with Programming Videocited by 1
  10. 2026A Concise Agent is Less Expert: Revealing Side Effects of Using Style Features on Conversational Agentscited by 1
  11. 2026IceBreaker for Conversational Agents: Breaking the First-Message Barrier with Personalized Starterscited by 1
  12. 2026Bye Bye Perspective API: Lessons for Building and Governing Measurement Infrastructurecited by 1
3 more
  1. 2026Mi:dm K 2.5 Procited by 0
  2. 2026Bridging Passive and Active: Enhancing Conversation Starter Recommendation via Active Expression Modelingcited by 0
  3. 2026kNNGuard: Turning LLM Hidden Activations into a Training-Free Configurable Guardrailcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025ConsistentChat: Building Skeleton-Guided Consistent Multi-Turn Dialogues for Large Language Models from Scratch
  2. 2024LAB: Large-Scale Alignment for ChatBots
  3. 2025LLM-Assisted Automated Deductive Coding of Dialogue Data: Leveraging Dialogue-Specific Characteristics to Enhance Contextual Understanding
  4. 2024Mixture-of-Instructions: Aligning Large Language Models via Mixture Prompting
  5. 2026Learning to Reason and Use Tools through Unsupervised Fine-Tuning in Task-Oriented Dialog Systems
  6. 2025CoDAE: Adapting Large Language Models for Education via Chain-of-Thought Data Augmentation
  7. 2025Towards Alignment-Centric Paradigm: A Survey of Instruction Tuning in Large Language Models
  8. 2024Inductive-Deductive Strategy Reuse for Multi-Turn Instructional Dialogues