Search papers, labs, and topics across Lattice.
This paper introduces an entity-only external interface that leverages task-conditioned relation materialization during inference to address the limitations of large software repositories in model context. By implementing a two-layer index that separates global routing from local entity focus, the authors evaluate their approach on DeepSeek-V4-Flash and SWE-bench Verified, achieving notable success rates of 92.1%, 94.2%, and 95.6% for base, one-layer, and two-layer conditions, respectively. These results highlight the effectiveness of the proposed method in enhancing code reasoning without the burden of maintaining explicit relation graphs.
Achieving up to 95.6% success in code reasoning without pre-built entity-relation edges could revolutionize how we manage large software repositories.
Large software repositories are often beyond model context limits. Training repository knowledge into models is costly and quickly stale, while local retrieval can miss scattered requirements, and explicit relation graphs add ongoing maintenance burden. We propose an entity-only external interface with task-conditioned relation materialization during inference. A two-layer index separates global routing from local entity focus and is evaluated on DeepSeek-V4-Flash and SWE-bench Verified. The base, one-layer, and two-layer conditions achieve 92.1%, 94.2%, and 95.6% success, respectively, under zero pre-built entity-relation edges.
