Daniel Khashabi

LLM agents can achieve high task accuracy while exhibiting profound epistemic *im*humility, failing to acknowledge uncertainty even when presented with contradictory evidence.

Modifying executable agent harnesses rather than model weights allows test-time adaptation to generalize across unseen reasoning tasks without a single gradient step.

A controlled cross-modality framework that instantiates the same task suite in a variety of modalities to test the Convergent Emergence Hypothesis and shows that paired-mapping ICL emerges across six modalities, surpasses controlled baselines, and has correlated per-task effects across five of them.

Together, these results show that composing complementary mechanisms substantially improves long-horizon memorization beyond what any individual mechanism achieves.

Models trained with VBVR-Pro not only excel in a controlled task space but also show significant transferability to external benchmarks, revealing critical insights into visual reasoning mechanisms.

Training LLMs with long contexts can paradoxically weaken their ability to retain knowledge, leading to poorer performance when context is not available.

Programmatic skill learning can slash agent costs while enhancing performance, with SpeedRunner leading the charge in cost-efficient adaptation.

SelfCompact reveals that language models can autonomously manage context decay, achieving up to 18.1 points improvement in performance while cutting token costs by 30-70%.

Trust functions can transform weak supervision into a powerful training signal, enabling models to achieve near-lossless generalization even with unreliable labels.

Agents struggle to match human performance in long-horizon tasks, revealing critical gaps in their learning capabilities during deployment.

Even state-of-the-art LLMs struggle to follow complex instruction hierarchies, achieving only ~40% accuracy when navigating conflicts across a dozen privilege levels in agentic tasks.

A 1000x larger video reasoning dataset reveals early signs of emergent generalization, offering a new foundation for training and evaluating spatiotemporal AI.