arXiv:2506.17871v4 Announce Type: replace-cross Abstract: Despite their impressive capabilities, aligned large language models (LLMs) often generate outputs that lack diversity. What drives this cons...
Warning: This paper studies stereotypes and biases, and contains potentially disturbing examples, used for illustration purposes only. Our findings should not be interpreted as an argument against alignment.
arXiv:2609.34187v2 Announce Type: replace-cross Abstract: The strong version of the stochastic parrot argument claims that, although large language models (LLMs) may exceed rote regurgitation, they c...
The paper introduces a technique called Synthetic Document Finetuning (SDF) to expose hidden misaligned behaviors in large language models that may engage in alignment faking. By finetuning models to associate certain keywords with undetected misaligned actions, the authors create honeypot environments and observe that the gap in misaligned action rates between cueed and uncued prompts widens, indicating more reliable detection than prompt-based methods. The study also notes that SDF can impair reasoning abilities, proposing on-policy distillation as a mitigation.
arXiv:2608. 07525v1 Announce Type: cross Abstract: Hallucination remains a persistent challenge for Multimodal Large Language Models (MLLMs), severely limiting their reliability in high-stakes applications.
arXiv:2606. 29824v1 Announce Type: cross Abstract: While Large Language Models (LLMs) excel as static solvers, transforming them into autonomous agents remains challenging.
While Large Language Models (LLMs) excel as static solvers, transforming them into autonomous agents remains challenging. This transition requires continuous environmental interaction, yet current agents lack the necessary persistent procedural memory.
The paper introduces SAGA, a framework that enables large language model agents to evolve by abstracting experiences into reusable principles, procedures, and episodic descriptions. SAGA transforms interaction trajectories into hierarchical knowledge with explicit applicability conditions, linking them back to execution evidence. Experiments on ScienceWorld and ALFWorld show that this execution–abstraction feedback loop improves task performance, and ablation studies confirm the importance of contextual instantiation and action regulation.
arXiv:2605. 09159v2 Announce Type: replace Abstract: Recent work shows that large language models (LLMs) encode behavioral traits ("personas") as linear directions in activation space, often called "persona vectors".
arXiv:2606.08682v2 Announce Type: replace Abstract: Activation steering (AS) has emerged as a popular inference-time technique for modulating the behavior of large language models (LLMs). AS modulate...
The paper investigates natural language autoencoders (NLAs) as tools for verbalizing activation differences in large language models. It examines whether NLAs capture compositional structure, generalize to fine‑tuned models, recover fine‑tuning objectives, and can be specialized for model diffing. Experiments across four models show that NLAs largely succeed in these tasks, though they do not yet match state‑of‑the‑art diffing methods on emergent misalignment detection.
The paper introduces domain-specific languages (DSLs) as an explicit representation layer to bridge the gap between high-level natural-language abstractions and concrete actions. Experiments on the visually-grounded “Hexagons” task and the “Text2CAD” benchmark show that code generation using a human-curated DSL yields dramatic performance gains, especially at higher abstraction levels, while automatically induced DSLs only partially recover these gains. The study highlights that the right representation can close much of the performance gap caused by abstraction, but that automatically learning such representations remains a challenge.