---
title: "面向QANTA 2026的多模态问答代理：基于置信度校准与增量推理的任务特定设计"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.09623v1"
permalink: "https://feed.astralor.com/items/itm_a350b3a62443409b38f80db0"
published: "2026-07-10T17:22:49.000Z"
score: 77
channel: "finance"
tags: ["GPT-4o","GPT-4o mini","任务特定代理","多模态问答","置信度校准"]
---

# 面向QANTA 2026的多模态问答代理：基于置信度校准与增量推理的任务特定设计

## AI 点评

> 这篇论文虽然只聚焦一个竞赛，但它的双代理架构设计非常巧妙：Tossup和Bonus分别应对不同优化目标，而且完全不依赖搜索和集成，仅靠推理策略+置信度校准就拿了第一。对做多模态问答、Agent、或者资源受限AI系统设计的人都很有启发。我尤其欣赏他们在数值推理上专门做了防过度自信设计，这点在实际产品中常被忽略。

## 速览

本文提交至ICML 2026高效多模态问答研讨会举办的QANTA 2026共享挑战赛。Quanta评估多模态问答系统，要求从逐步揭示的文本和图像中回答金字塔式问题，并满足现实效率约束。挑战包含两任务：Tossup（不确定性下决定何时作答）与Bonus（精确答案选择及人类采用）。作者构建了任务特定双代理架构：Tossup代理使用GPT-4o-mini级模型，结合置信度校准回答和领域特定数值推理策略，减少孤立数值线索导致的过度自信；Bonus代理使用GPT-4o级模型，融入导语感知推理、结构化关系推理和多模态证据整合，提升精确答案选择。系统不依赖检索流水线或模型集成，仅使用托管环境内的推理策略和置信度校准，最终以0.402总榜最高分（Tossup 0.238，Bonus效果0.164）证明轻量级任务特定推理策略在资源受限多模态问答基准上的有效性。

## 关键点

- 提出任务特定双代理架构，分别处理Tossup（不确定作答时机）与Bonus（精确答案选择）任务。
- Tossup代理采用置信度校准与领域特定数值推理，减少孤立数值线索的过度自信。
- Bonus代理融合导语感知推理、结构化关系推理与多模态证据整合。
- 系统无需检索或模型集成，在托管环境下以0.402总分获得排行榜最高成绩。

## 全文

We present our submission to the QANTA 2026 shared challenge at the ICML 2026 Workshop on Efficient Multimodal Question Answering (EMM-QA). Quanta evaluates multimodal quizbowl systems that answer pyramid-style questions from incrementally revealed text and accompanying images while operating under realistic efficiency constraints. The challenge consists of two distinct tasks: Tossup questions, which require deciding when to answer under uncertainty, and Bonus questions, which emphasize accurate answer selection and human adoption. To address these differing objectives, we develop a task-specific two-agent architecture. Our Tossup agent utilizes a GPT-4o-mini-class model (referred to as GPT-4.1-mini in the competition logs) with confidence-calibrated answering and a domain-specific numeric reasoning policy that reduces overconfident predictions from isolated quantitative clues. Our Bonus agent uses GPT-4o-class model (referred to as GPT-4.1) with leadin-aware reasoning, structured relational reasoning, and multimodal evidence integration to improve exact answer selection. Rather than relying on a retrieval pipeline or model ensembles, our approach emphasizes efficient reasoning policies and confidence calibration within a hosted-only environment. Our system achieved the highest overall leaderboard score of 0.402, including a Tossup score of 0.238 and a Bonus Effect score of 0.164. The results demonstrate that lightweight, task-specific reasoning strategies can provide strong performance on resource-constrained multimodal question answering benchmarks.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_a350b3a62443409b38f80db0