We introduce \(\infty\)-THOR, a new framework for long-horizon embodied tasks that advances long-context understanding in embodied AI. \(\infty\)-THOR provides:
(1) a generation framework for synthesizing scalable, reproducible, and unlimited long-horizon trajectories; (2) a novel embodied QA task, Needle(s) in the Embodied Haystack, where multiple scattered clues across extended trajectories test agents' long-context reasoning ability; and (3) a long-horizon dataset and benchmark suite featuring complex tasks that span hundreds of environment steps, each paired with ground-truth action sequences. To enable this capability, we explore architectural adaptations, including interleaved Goal-State-Action modeling, context extension techniques, and Context Parallelism, to equip LLM-based agents for extreme long-context reasoning and interaction. Experimental results and analyses highlight the challenges posed by our benchmark and provide insights into training strategies and model behaviors under long-horizon conditions. Our work provides a foundation for the next generation of embodied AI systems capable of robust, long-term reasoning and planning.
Our generation framework can generate unlimited tasks, the trajectories can be exceptionally long, exceeding 1M context tokens or beyond.
\(\infty\)-THOR introduces a new challenging task, Needle(s) in the Embodied Haystack (NiEH). Unlike the standard Needle in a Haystack task, which focuses on recalling a single clue in text, NiEH poses two main challenges: (1) multiple scattered clues (Needles) and (2) multi-modal inputs that combine visual and linguistic observations from the environment (Embodiment). This task is designed to evaluate the agent's ability to recall and reason about previously encountered environmental details, such as identifying objects and recalling performed actions. Figure 3 and 4 present examples of the two NiEH task types. In the single-evidence setting, a question is answerable based on a single observation step; in the multi-evidence setting, multiple temporally distant steps must be combined to answer the question.
Our benchmark uniquely features tasks with a synthetic final goal, which involves multiple objects that appear at distant time steps, requiring multi-step reasoning across over hundreds of steps. Figure 3 illustrates an example: the agent observes the tomato at an early step (t=17) and the counter top much later (t=560). Then, the final task is given at t=670, which requires the agent to place the tomato on the counter top. This setup highlights the challenge of long-horizon dependency, where key objects and locations must be remembered and acted upon after hundreds of steps.