让视觉推理模型逐步学习任务变换过程，提升抽象思维能力。
TraceViT: Grounded Trace Supervision for Visual Abstract Reasoning
- 用逐步变换的中间状态监督模型迭代过程
- 在ARC-AGI-1上达到67.8%的pass@2准确率
- 适合研究可解释视觉推理与多步逻辑建模的学者
抽象与推理语料库（ARC）测试模型能否从少量输入输出样例中推断出未见的变换并应用于新网格。循环视觉推理器通过多轮迭代优化预测，但传统训练仅约束最终输出，中间过程缺乏监督。我们提出应让中间迭代逐步遵循变换路径。为此引入TraceViT，一种基于语义单调变换链的循环视觉推理模型。这些链通过重写并验证程序化任务实现得到，将每个解分解为中间网格状态。每轮迭代由少量示范中的任务参考和当前网格状态的对象工作区共同引导。由于链长可能与循环次数不一致，采用软对齐仅强制顺序，允许模型自由分配迭代次数。TraceViT在ARC-AGI-1上取得67.8% pass@2，在ARC-AGI-2上取得24.3% pass@2。在ARC-AGI-1上的受控消融实验表明，仅有接地机制时，轨迹监督才真正有效。代码与数据将在https://github.com/LiuBinnan/TraceViT公开。
原文摘要 · Abstract (English)
The Abstraction and Reasoning Corpus (ARC) tests whether a model can infer an unseen transformation from a few input-output examples and apply it to a new grid. Looped visual reasoners refine predictions over multiple iterations, but conventional training constrains only the final output, leaving intermediate refinements unconstrained. We propose that these refinements should instead follow the transformation step by step. We introduce TraceViT, a looped visual reasoner trained with semantically monotonic transformation chains. We obtain these chains by rewriting and verifying programmatic task implementations, decomposing each solution into intermediate grid states. Each iteration is grounded by a task reference derived from the few-shot demonstrations and an object workspace representing the current grid state. Because these chains may differ in length from the loop, soft trace alignment enforces only their ordering, letting the model allocate iterations freely. TraceViT achieves 67.8% pass@2 on ARC-AGI-1 and 24.3% on ARC-AGI-2. Controlled ablations on ARC-AGI-1 show that trace supervision becomes beneficial only when paired with grounding. Code and data will be available at https://github.com/LiuBinnan/TraceViT.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。