Tag
The paper proposes a heterogeneity-driven framework for selecting complementary LLM teams by profiling error decorrelation and predictive divergence, then greedily optimizing a quality–complementarity objective, showing gains over quality-only top-k baselines across multiple benchmarks.
CoMed introduces a post-anchor controller for selective cross-model collaboration in multi-LLM systems, improving accuracy on benchmarks like MedQA and HLE while reducing computational costs compared to dense collaboration.
AIRI is an AI companion that features real-time voice chat, gaze-tracking avatars, memory retention, and gameplay support for Minecraft and Factorio, with compatibility for over 30 LLMs and local inference via WebGPU.
This paper proposes MIDAS, a multi-LLM framework for data-adaptive summarization that automates prompt optimization for domain-specific enterprise use cases, achieving strong improvements over prior methods on customer ticket summarization benchmarks.
Opti-Q is a database-inspired optimizer for multi-LLM question answering that plans execution DAGs to optimize answer quality under cost, latency, and energy constraints, achieving significant improvements on benchmarks.
This paper introduces an uncertainty-aware trust estimation method for aggregating predictions from multiple LLMs, adapting structured expert judgment with Cooke-style log weighting to penalize overconfident incorrect predictions. Evaluations on MMLU and MMLU-Pro show that this approach achieves superior accuracy-reliability balance under heterogeneous and contaminated expert panels.
This paper presents SynthAVE, a large-scale human-validated benchmark for attribute value extraction in e-commerce, using a multi-LLM arena framework with 21 judge configurations to validate synthetic labels efficiently and cost-effectively while maintaining quality parity with human review.
A detailed guide on building an agentic research framework using a multi-LLM system with persistent memory, allowing researchers to avoid re-explaining context across sessions by leveraging file-based identity, project docs, and memory indices.