dingxingdi/paper_fast_search_backup/information-asymmetric-theory-of-mind-reasoning

Skill: information-asymmetric-theory-of-mind-reasoning

1. Capability Definition & Real Case

• Professional Definition: The ability to reason about who knows what, and what a specific interlocutor would believe, when conversation participants have unequal access to information due to joining, leaving, or otherwise missing parts of a dialogue.
• Dimension Hierarchy: Social and Persona Modeling->Mental-state Reasoning->information-asymmetric-theory-of-mind-reasoning

Real Case

[Case 1]

• Initial Environment: A group chat begins with Linda and David discussing Linda’s new dog. Kailey leaves to get coffee during the part where the breed is mentioned and returns only after the group shifts to talking about the dog’s tricks.
• Real Question: What breed would Kailey think Linda’s dog is?
• Real Trajectory: 1. Track which utterances occurred before Kailey left, during her absence, and after she returned. 2. Mark the dog-breed detail as inaccessible to Kailey. 3. Distinguish omniscient-view truth from Kailey-centric accessible knowledge. 4. Answer from Kailey’s perspective rather than from the full transcript.
• Real Answer: Kailey does not know the breed.
• Why this demonstrates the capability: The transcript contains the true breed, so a naive omniscient assistant can answer incorrectly by collapsing its own knowledge with the absent character’s mental state. Success requires modeling information accessibility and belief from a constrained perspective inside a live conversation.

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