LLMs
Latest breakthroughs in Large Language Models
Articles
Auto-FL-Research: Agentic Search for Federated Learning Algorithms⭐8
Auto-FL-Research introduces an agentic search framework for automating federated learning algorithm design. Tests on 11 healthcare and LEAF benchmarks show perf...
PACE: A Neuro-Symbolic Framework for Plausible and Actionable Counterfactual Explanations⭐7
PACE proposes a neuro-symbolic framework for counterfactual explanations, combining neural prediction with symbolic constraints to generate more realistic and a...
Bounded Morality: Defining the Space of Moral Computation⭐8
This paper proposes Bounded Morality, a framework analyzing moral decision-making under computational constraints, defining moral breadth and depth tradeoffs fo...
Constructive Alignment: Governing Preference Dynamics in Human-AI Interaction⭐8
Constructive Alignment reframes AI alignment as governing dynamic human preference trajectories to ensure coherence, reflectively endorsed values, and resistanc...
Contrastive Reflection for Iterative Prompt Optimization⭐9
Contrastive Reflection improves agentic IR prompts by comparing failed vs. successful traces, boosting HotpotQA accuracy from 51.4% to 60.4%.
What Drives Interactive Improvement from Feedback?⭐9
Multi-turn AI agents often improve from retrying, not from feedback quality. Gains depend more on the student’s ability to use feedback than on the teacher’s ex...
Recursive Self-Evolving Agents via Held-Out Selection⭐8
RSEA enables LLM agents to self-improve via recursive context evolution without weight updates, using held-out selection to prevent regression and ensure safe t...
When Does Personality Composition Matter for Multi-Agent LLM Teams?⭐9
Personality composition impacts multi-agent LLM teams differently by task: low agreeableness hinders open-ended tasks but has minimal effect on structured codin...
The AI-Model Network: Concept, Current State, and Future Directions⭐8
The AI-Model Network proposes a global system for interconnecting, sharing, and collaborating across heterogeneous AI models, addressing high costs and deployme...
Life After Benchmark Saturation: A Case Study of CORE-Bench⭐8
Learn what happens after AI benchmarks hit accuracy saturation. A CORE-Bench case study reveals six crucial dimensions beyond accuracy.