Mirella Lapata

Pixel-level table compression can dramatically reduce token usage while enhancing accuracy in document question answering, challenging conventional methods.

Efficiently routing queries in AI systems can be achieved without the costly overhead of exhaustive value estimation, thanks to novel policies that balance accuracy and cost.

State-of-the-art audio description systems struggle to match expert human performance, revealing significant gaps in current methodologies.

Literary character analysis gets a boost: representing novels as dynamic graphs that fuse character interactions with their textual context unlocks better performance on character-related tasks.

Disabling LLMs' built-in reasoning can paradoxically improve character description generation, but the right kind of external reasoning (QA-guided) boosts performance even further.

Ditch the slow, error-prone tool calls: Pearl learns to reason with multimodal tools entirely in the latent space, matching or exceeding SOTA performance without ever explicitly invoking a tool at inference time.