The authors formulate per-query harness construction as a tree search over actions that define agents, instructions, and tools.
System that builds agent teams per query outperforms all baselines
SHIFT uses a local LLM to predict harness utility without executing candidate designs, selecting agents, instructions, and tools for each task.
Academic
Som Sagar · Shasha Li · Hejie Cui · Ransalu Senanayake · Sercan Ö. Arık
Arizona State University
Research Digest··3 min read
The authors introduce SHIFT, a framework that constructs a multi-agent harness—the combination of agent roles, instructions, and tools—for each query without executing candidate designs during search.
Why this paper
From Arizona State University
In one line
A learned architect predicts harness utility, enabling per-query multi-agent system construction without executing candidates during search.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ✓Reports numbers on named benchmarks (3 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§