Fewer Clarifications, Better Code: Benchmarking Cross-Session Personalized Ambiguity Adaptation in Coding Assistants

2026-07-29

AI Summary

This work addresses the recurring issue of personalized ambiguity in cross-session programming by introducing the first Cross-session Ambiguity Personalization Adaptation (CAPA) task. CAPA leverages a user’s historical resolved sessions to identify and resolve repeated ambiguities in new requests, thereby avoiding redundant clarifications. The authors construct a benchmark dataset encompassing six distinct ambiguity mechanisms and propose a three-stage controllable generation pipeline augmented with a user-history gating mechanism to enhance inference efficiency. Evaluation across 600 sessions demonstrates that incorporating user history significantly improves the success rate of large language models in generating correct code on the first attempt and effectively reduces the number of required interaction rounds.

Abstract

AI-assisted coding increasingly translates informal user intent into executable software, yet coding requests often contain ambiguities that recur in user-specific ways across tasks and sessions. Existing disambiguation methods typically address each ambiguous request in isolation within the current coding session, often through eliciting additional clarification. However, whether resolved session history from the same user can serve as memory for resolving recurring personalized ambiguity in a newly opened session remains underexplored. We formulate personalized ambiguity adaptation as a new task: given a user's previously resolved coding sessions and a new ambiguous request, an assistant should identify the recurring ambiguity pattern, produce the intended executable solution, and minimize clarification. To benchmark this task, we introduce CAPA, which characterizes personalized coding ambiguity through six mechanisms and injects these mechanisms into unambiguous executable tasks using a controlled three-stage generation pipeline. CAPA contains 600 coding sessions across 60 balanced user--ambiguity cells, including 300 held-out evaluation sessions. We evaluate 12 recent LLMs under no-history and same-user-history conditions using executable success, first-turn success, and turns-to-completion. Our analyses examine task difficulty, user identity, and memory-based history use, and we further propose same-user history gating as a lightweight inference-time method. CAPA provides a foundation for developing long-term coding assistants that better align generated code with user intent while reducing repeated clarification.

Problem

• personalized ambiguity
• cross-session adaptation
• coding assistants
• disambiguation
• user intent

Innovation

• personalized ambiguity adaptation
• cross-session memory
• coding assistants
• CAPA benchmark
• history gating