TactiNavi
An LLM-based game coaching service system deployed in a commercial open-world game. It models strategy applicability as explicit preconditions and spans the full pipeline of offline compilation, online execution, fault diagnosis and release.
Timeline (last 90 days)
TactiNavi is an LLM coaching service system deployed in a commercial open-world game.
The core innovation of TactiNavi is treating the applicability conditions (prerequisites) of strategies as first-class citizens, running through the entire pipeline of offline compilation, online execution, fault diagnosis, and release.
TactiNavi's offline compiler transforms versioned game evidence and expert guidance into strategy packages with explicit preconditions.
TactiNavi's service plane treats preconditions as three-valued logic conditions, evaluates them on structured player state, and records decision traces.
TactiNavi's control plane leverages condition-rule dependencies to validate revisions with held-out traffic before release.
In TactiNavi, three-valued logic allows the system to mark conditions that are missing, stale, or undecidable as "unknown", and then adopt conservative strategies (such as not activating a strategy, downgrading suggestions, or requesting more information).
In a retrospective evaluation covering three deployment cycles and a total of 1,122 recorded coaching dialogues, conditional strategy packages raised the correctness rate from 57.9% to 82.0% (compared with atomic facts).
In the above retrospective evaluation, conditional strategy packages reduced strategy misuse caused by changes in player state from 18.1% to 10.4%.
TactiNavi's trace-based claim verification achieved 86.4% precision at a median response latency of 2.07 seconds.
When re-run without any changes to the system, 38.4% of failed dialogues recovered.
2 more timeline events
All Facts (12)
TactiNavi is an LLM coaching service system deployed in a commercial open-world game.
50%UnverifiedThe core innovation of TactiNavi is treating the applicability conditions (prerequisites) of strategies as first-class citizens, running through the entire pipeline of offline compilation, online execution, fault diagnosis, and release.
50%UnverifiedTactiNavi's offline compiler transforms versioned game evidence and expert guidance into strategy packages with explicit preconditions.
50%UnverifiedTactiNavi's service plane treats preconditions as three-valued logic conditions, evaluates them on structured player state, and records decision traces.
50%UnverifiedTactiNavi's control plane leverages condition-rule dependencies to validate revisions with held-out traffic before release.
50%UnverifiedIn TactiNavi, three-valued logic allows the system to mark conditions that are missing, stale, or undecidable as "unknown", and then adopt conservative strategies (such as not activating a strategy, downgrading suggestions, or requesting more information).
50%UnverifiedIn a retrospective evaluation covering three deployment cycles and a total of 1,122 recorded coaching dialogues, conditional strategy packages raised the correctness rate from 57.9% to 82.0% (compared with atomic facts).
50%UnverifiedIn the above retrospective evaluation, conditional strategy packages reduced strategy misuse caused by changes in player state from 18.1% to 10.4%.
50%UnverifiedTactiNavi's trace-based claim verification achieved 86.4% precision at a median response latency of 2.07 seconds.
50%UnverifiedWhen re-run without any changes to the system, 38.4% of failed dialogues recovered.
50%UnverifiedIn the retrospective check, only 1 of 15 released revision units passed the release check.
50%UnverifiedTactiNavi's practice distills three lessons: treat applicability as a first-class service object; design the granularity and scale of traces appropriately for downstream diagnosis; and build a null model into the release process.
50%