9 papers
CLIMB: A Clinical Multimorbidity Benchmark for Diagnosing Co-occurring Conditions through Multiturn Conversations
Yusuf Kesmen, Aniruddha Mukherjee, Yena Chang +7
Patients often have several co-occurring clinical conditions, and the findings needed to identify and disambiguate them emerge over the course of a consultation. Evaluating clinica…
The Working Set of a Coding Agent: Coherence Debt in Repository-Scale Tasks
Bardia Mohammadi, Lars Klein, Aman Chadha +2
Repository-scale coding requires an agent to keep tests, imports, configuration, and migration rules consistent within a bounded context window. We model this as reconstructing a c…
Ghost Tool Calls: Issue-Time Privacy for Speculative Agent Tools
Bardia Mohammadi, Lars Klein, Akhil Arora +1
Tool-augmented language agents speculatively issue likely future tool calls to hide latency, but those calls leak inferred user intent to external services before the agent commits…
MoBayes: A Modular Bayesian Framework for Separating Reasoning from Language in Conversational Clinical Decision Support
Yusuf Kesmen, Fay Elhassan, Jiayi Ma +7
Large language models (LLMs) are increasingly used for conversational clinical decision support, yet they conflate next token prediction with probabilistic decision making. We argu…
Agentic Proving for Program Verification
Alessandro Sosso, Akhil Arora, Bas Spitters
Agentic systems have recently emerged as state-of-the-art approaches for automated theorem proving in formal mathematics. To assess how far these capabilities extend to program ver…
Atomix: Timely, Transactional Tool Use for Reliable Agentic Workflows
Bardia Mohammadi, Nearchos Potamitis, Lars Klein +2
LLM agents execute multi-step workflows that mutate external state through tools. Common orchestrators treat tool return as the settlement trigger, so faults, speculation, and conc…
DIVERGE: Diversity-Enhanced RAG for Open-Ended Information Seeking
Tianyi Hu, Niket Tandon, Akhil Arora
Existing retrieval-augmented generation (RAG) systems often assume that each query has a single correct answer. This assumption overlooks open-ended information-seeking scenarios w…
ReasonBENCH: Benchmarking the (In)Stability of LLM Reasoning
Nearchos Potamitis, Vansh Ramani, Har Ashish Arora +3
Benchmark scores for LLM reasoning systems are reported as single numbers, yet the same model, strategy, and task can produce meaningfully different answers and costs across repeat…
Participatory AI: A Scandinavian Approach to Human-Centered AI
Niklas Elmqvist, Eve Hoggan, Hans-Jörg Schulz +45
AI's transformative impact on work, education, and everyday life makes it as much a political artifact as a technological one. Current AI models are opaque, centralized, and overly…