dingxingdi/paper_fast_search_backup/episodic-and-multi-hop-memory-reasoning

Skill: episodic-and-multi-hop-memory-reasoning

1. Capability Definition & Real Case

• Professional Definition: The capability to answer questions whose evidence is distributed across distant segments, days, or episodes, requiring retrieval of multiple memories and a reasoning chain that links them into one conclusion.
• Dimension Hierarchy: Long-Horizon Reasoning->Global Content Integration->episodic-and-multi-hop-memory-reasoning

Real Case

[Case 1]

• Initial Environment: A week-long egocentric video log with hierarchical summaries and retrievable subclips. The relevant supermarket escalator event occurred days before the question timestamp.
• Real Question: Who was in front of me last time I stood on the escalator in the supermarket?
• Real Trajectory: Retrieve the week-level memory using 'escalator' and 'supermarket', drill down to the matching day, inspect the identified escalator subclip, and connect the retrieved high-level memory about companions with the frame-level visual evidence.
• Real Answer: Tasha.
• Why this demonstrates the capability: No single local clip is enough unless the agent first finds the right day and the right escalator event. The problem therefore tests long-horizon episodic retrieval plus multi-hop reasoning over summary-level and frame-level evidence.

Pipeline Execution Instructions

To synthesize data for this capability, you must strictly follow a 3-phase pipeline. Do not hallucinate steps. Read the corresponding reference file for each phase sequentially:

1. Phase 1: Environment Exploration

Read the exploration guidelines to discover raw knowledge seeds: references/EXPLORATION.md

1. Phase 2: Trajectory Selection

Once Phase 1 is complete, read the selection criteria to evaluate the trajectory: references/SELECTION.md

1. Phase 3: Data Synthesis

Once a trajectory passes Phase 2, read the synthesis instructions to generate the final data: references/SYNTHESIS.md