analiza aiqbench
This research introduces SAR, a lightweight LoRA adapter enabling fine-tuned models to reliably self-report hidden behaviors without external probes or test data. By significantly reducing hallucinations compared to existing introspection methods, the technique offers practitioners a robust tool for auditing model alignment and verifying what was actually learned during training. This advancement directly addresses critical challenges in AI safety and interpretability by making internal model states more transparent and trustworthy.
Revealing Hidden Model Behaviors with Task-Specific Self-Reports
arXiv:2607.03640v2 Announce Type: replace-cross Abstract: Fine-tuning can give a language model a hidden behavior--it may give false answers under a narrow condition, or give harmful advice only when a prompt touches a particular topic. We introduce the Stabilized Adapter for self-Report (SAR), a lightweight LoRA adapter that makes a fine-tuned model describe its own hidden behavior in plain language, using only the model and the dataset it was trained on. Across seven implanted behaviors, SAR detects the hidden behavior in every one--even when the model has generalized into broad misalignment that the training data alone does not predict. Introspection Adapters (IA), the closest existing baseline, detects some behaviors from our suite but misses others entirely--and where it misses, it hallucinates, consistently reporting wrong behaviors. SAR retains positive signal on every setting where IA fails and roughly halves the rate of hallucinations. This gives practitioners a more reliable tool to audit a fine-tuned model and answer ``what did it actually learn?'' type of questions.
Abstract:Fine-tuning can give a language model a hidden behavior--it may give false answers under a narrow condition, or give harmful advice only when a prompt touches a particular topic. We introduce the Stabilized Adapter for self-Report (SAR), a lightweight LoRA adapter that makes a fine-tuned model describe its own hidden behavior in plain language, using only the model and the dataset it was trained on. Across seven implanted behaviors, SAR detects the hidden behavior in every one--even when the model has generalized into broad misalignment that the training data alone does not predict. Introspection Adapters (IA), the closest existing baseline, detects some behaviors from our suite but misses others entirely--and where it misses, it hallucinates, consistently reporting wrong behaviors. SAR retains positive signal on every setting where IA fails and roughly halves the rate of hallucinations. This gives practitioners a more reliable tool to audit a fine-tuned model and answer ``what did it actually learn?'' type of questions.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2607.03640 [cs.CL]
(or arXiv:2607.03640v2 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2607.03640 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 257 cuvinte