Search Beyond What Can Be Taught: Evolving the Knowledge Boundary in Agentic Visual Generation

• URL: http://arxiv.org/abs/2607.05382v3
• Date: Thu, 09 Jul 2026 06:57:10 GMT
• Title: Search Beyond What Can Be Taught: Evolving the Knowledge Boundary in Agentic Visual Generation
• Abstract summary: Visual generators excel at rendering, but they confidently fabricate what they do not know.<n>We construct SearchGen-20K and SearchGen-Bench, with 20,839 prompts spanning twelve failure categories and twenty-two domains.<n>On SearchGen-Bench, frontier open generators score only 21 to 28 out of 100, a 40-point collapse invisible to existing benchmarks.
• Score: 83.53893718885796
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Visual generators excel at rendering, but they confidently fabricate what they do not know. User requests are unbounded, evolving, and deeply long-tailed: new characters, trending entities, post-cutoff events, and more. This world-knowledge bottleneck is structural: generators are trained on fixed corpora, but the visual world is open-ended. We construct SearchGen-20K and SearchGen-Bench, with 20,839 prompts spanning twelve failure categories and twenty-two domains, paired with a pre-executed multimodal SearchGen-Corpus-1M to support offline, reproducible research. On SearchGen-Bench, frontier open generators score only 21 to 28 out of 100, a 40-point collapse invisible to existing benchmarks. The natural remedy is to employ search tools, enabling agentic visual generation. However, we find that naive search fails: it retrieves indiscriminately, injecting noise into prompts the generator already handles. We trace the root cause to a generator-specific, evolving knowledge boundary: the divide between what a generator can internalize through training and what must remain in external context. Although this boundary is hard to specify in advance, we show that it is discoverable through a teach-then-search co-training framework. Even a minimal version of this co-training recipe produces monotonic improvement, laying the foundation for recursive self-improvement in visual generation that can meet world-knowledge-grounded requests. We release the full dataset, co-training corpus, and search corpus as a replayable harness for tool-augmented, world-knowledge-grounded visual generation.

Related papers

• Procedural Content Metageneration via Program Search and Continual Abstraction Discovery [3.899863694644006]
  We introduce Continual Abstraction Discovery, or CAD, which extracts reusable primitives from high-fitness programs into a run-specific helper module.<n>We study this approach in Sokoban, Zelda, Dangerous Dave, and Lode Runner.
  arXiv  Detail & Related papers  (2026-08-18T16:03:22Z)
• G-ReAct: Graph-Guided Deep Search via Structure-State Co-Evolution [16.818086150132853]
  $textbfG-ReAct is a reasoning framework for deep search.<n>It organizes reasoning as $textbfstate evolution over a fixed-topology query graph.<n>G-ReAct supports both training and inference.
  arXiv  Detail & Related papers  (2026-08-02T15:43:37Z)
• SearchEyes: Towards Frontier Multimodal Deep Search Intelligence via Search World Simulation [26.8807904319691]
  We present textbfSearchEyes, which uses a typed knowledge graph as the backbone of a emphsimulated search world<n>We propose textbfPerception-Knowledge Chains (PKC) to sample constrained multi-hop paths over the visual-knowledge intersection of Wikidata5M.<n>Experiments on six multimodal knowledge-intensive benchmarks show that SearchEyes achieves state-of-the-art performance among open-source multimodal search agents.
  arXiv  Detail & Related papers  (2026-07-07T07:43:04Z)
• Thinking-while-Generating: Interleaving Textual Reasoning throughout Visual Generation [79.31152006811438]
  Thinking-while-Generating (TwiG) is the first interleaved framework that enables co-evolving textual reasoning throughout the visual generation process.<n>To unveil the potential of this framework, we investigate three candidate strategies, zero-shot prompting, supervised fine-tuning, and reinforcement learning.
  arXiv  Detail & Related papers  (2025-11-20T18:59:52Z)
• UniSearch: Rethinking Search System with a Unified Generative Architecture [20.448690421956023]
  UniSearch is a unified generative search framework for Kuaishou Search.<n>UniSearch replaces the cascaded pipeline with an end-to-end architecture that integrates a Search Generator and a Video Generator.<n>Extensive experiments on industrial-scale datasets, together with online A/B testing in both short-video and live search scenarios, demonstrate the strong effectiveness and deployment potential of UniSearch.
  arXiv  Detail & Related papers  (2025-09-08T17:08:26Z)
• Searching for the Fakes: Efficient Neural Architecture Search for General Face Forgery Detection [4.19882227319634]
  We develop an end-to-end framework based on neural architecture search (NAS) for deepfake detection. We propose a novel performance estimation metric, which guides the search process to select more general models. Compared with state-of-the-art networks artificially designed, our method achieves competitive performance in both in-dataset and cross-dataset scenarios.
  arXiv  Detail & Related papers  (2023-06-15T03:01:13Z)
• Enhancing Retrieval-Augmented Large Language Models with Iterative Retrieval-Generation Synergy [164.83371924650294]
  We show that strong performance can be achieved by a method we call Iter-RetGen, which synergizes retrieval and generation in an iterative manner. A model output shows what might be needed to finish a task, and thus provides an informative context for retrieving more relevant knowledge. Iter-RetGen processes all retrieved knowledge as a whole and largely preserves the flexibility in generation without structural constraints.
  arXiv  Detail & Related papers  (2023-05-24T16:17:36Z)
• Active Retrieval Augmented Generation [123.68874416084499]
  Augmenting large language models (LMs) by retrieving information from external knowledge resources is one promising solution. Most existing retrieval augmented LMs employ a retrieve-and-generate setup that only retrieves information once based on the input. We propose Forward-Looking Active REtrieval augmented generation (FLARE), a generic method which iteratively uses a prediction of the upcoming sentence to anticipate future content.
  arXiv  Detail & Related papers  (2023-05-11T17:13:40Z)
• Momentum Decoding: Open-ended Text Generation As Graph Exploration [49.812280360794894]
  Open-ended text generation with autoregressive language models (LMs) is one of the core tasks in natural language processing. We formulate open-ended text generation from a new perspective, i.e., we view it as an exploration process within a directed graph. We propose a novel decoding method -- textitmomentum decoding -- which encourages the LM to explore new nodes outside the current graph.
  arXiv  Detail & Related papers  (2022-12-05T11:16:47Z)
• Unsupervised Text Generation by Learning from Search [86.51619839836331]
  TGLS is a novel framework to unsupervised Text Generation by Learning. We demonstrate the effectiveness of TGLS on two real-world natural language generation tasks, paraphrase generation and text formalization.
  arXiv  Detail & Related papers  (2020-07-09T04:34:48Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.