This paper studies the phenomenon of behavioral couplings in large language models, where fine-tuning on one behavior induces another.
This paper introduces the concept of behavioral couplings in large language models, which is a novel finding that challenges the current understanding of model behavior.
Before reading this…
Applications
To understand this paper, make sure you know these concepts first:
Fine-tuning large language models (LLMs) can induce behaviors beyond those explicitly targeted. For example, training a model to respond more warmly can make it more sycophantic, and training it to write insecure code can cause broad misalignment on prompts unrelated to coding. We refer to a relationship between behaviors, in which fine-tuning on one brings out another, as a behavioral coupling. In this work, we study formation and properties of behavioral couplings. We find that when a model is trained on documents describing two distinct behaviors together, fine-tuning it to elicit one behavior causes the other to emerge as well. For instance, when training documents describe a fictitious character that both explains scientific concepts in a playful, rubber-duck-like manner and discourages human friendships, a model subsequently fine-tuned to explain science like a rubber duck also begins encouraging users to end their friendships, whereas a model that was not trained on these documents shows no such behavior after the same fine-tuning. These results show that behavioral couplings can be formed through document training and later surface as unintended behaviors. More concerningly, we find that couplings can compose. When two document sets share a behavior (document A pairs behaviors 1 and 2, and document B pairs behaviors 2 and 3), fine-tuning on behavior 1 elicits both behaviors 2 and 3. This can make it difficult to trace where an emergent misbehavior originated, since the coupled behaviors may never appear together in any single document. These findings suggest that preventing the unintended effects of later fine-tuning may require evaluation that goes beyond a model's current behavior.
AmbiHTMLBench: A Benchmark for HTML-based Visually Rich Document parsing
This paper introduces a benchmark for evaluating document parsing systems that o…
Stellar: Scalable Multimodal Document Retrieval for Natural Language Queries
This paper presents Stellar, a scalable multimodal document retrieval framework…
GLOBE: Evolving Structured Agentic Memory for E-commerce Product Optimization in Generative Search
This paper introduces GLOBE, an agentic framework for continual self-improvement…
Select or Rewrite: Oracle-Guided Fine-Tuning for Generative ASR Post-Editing from N-Best Hypotheses
The paper proposes oracle-guided fine-tuning to improve post-editing of ASR tran…
Beyond Behavioral Accuracy: Evaluating Individual-Level Fidelity in LLM Investor Simulation
This paper investigates whether large language models can faithfully reproduce a…
Listwise Cross-Encoder Fine-Tuning vs. Agentic Instruction Tuning for LLM Rerankers: A Systematic St…
This paper compares two reranking paradigms for health insurance information ret…
Maintaining IoT Device Identification under Concept Drift via Budget-Aware Traffic Labeling
This paper conducts a two-year study on IoT traffic to characterize behavioral e…
Two Conformal Constructions for Adaptive Within-Document AI-Text Screening
The paper presents two finite-sample constructions for controlling false alerts…