Litmus (Re)Agent: A Benchmark and Agentic System for Predictive Evaluation of Multilingual Models

TL;DR

This paper introduces Litmus (Re)Agent, a benchmark and agentic system for predicting multilingual model performance when direct evaluation data is unavailable, demonstrating improved inference in sparse evidence scenarios.

Contribution

It presents a new benchmark with 1,500 questions across six tasks and a novel agentic system that decomposes queries, retrieves evidence, and synthesizes predictions.

Findings

Litmus (Re)Agent outperforms six other systems overall.

Largest gains observed in transfer-heavy scenarios with weak evidence.

Structured agentic reasoning improves multilingual performance estimation.

Abstract

We study predictive multilingual evaluation: estimating how well a model will perform on a task in a target language when direct benchmark results are missing. This problem is common in multilingual deployment, where evaluation coverage is sparse and published evidence is uneven across languages, tasks, and model families. We introduce a controlled benchmark of 1,500 questions spanning six tasks and five evidence scenarios. The benchmark separates accessible evidence from ground truth, enabling evaluation of systems that must infer missing results from incomplete literature evidence. We also present Litmus (Re)Agent, a DAG-orchestrated agentic system that decomposes queries into hypotheses, retrieves evidence, and synthesises predictions through feature-aware aggregation. Across six systems, Litmus (Re)Agent achieves the best overall performance, with the largest gains in transfer-heavy scenarios where direct evidence is weak or absent. These results show that structured agentic reasoning is a promising approach to multilingual performance estimation under incomplete evidence.