Results | FIRE 2026 Shared Task

The final results of the FIRE 2026 Shared Task on AI4Industrial Automation are presented separately for the two subtasks. The rankings are based on the final evaluation score obtained on the official test data.

Evaluation Formula

The final score is calculated by comparing the generated program with the corresponding Gold Standard Code.

Final Score = 0.25 × Cosine Similarity + 0.25 × Structural Match Score + 0.50 × PDG Score

Metric	Weight
Cosine Similarity (CS)	25%
Structural Match Score (SMS)	25%
Program Dependence Graph Similarity (PDG)	50%

Note: The final ranking is determined separately for each subtask. A separate leaderboard is provided for Task A (Natural Language to Structured Text) and Task B (Natural Language to Rust Program).

Task A: Natural Language to Structured Text (ST)

The following leaderboard presents the results for systems that generate IEC 61131-3 Structured Text programs from natural language descriptions.

Rank	Team Name	Final Score
1	8 Bit Thugs	81.78
2	team meooo	79.31
3	Treygram	78.83
4	Master Mind	76.63
5	CSNLP	75.11
6	AutoLogic AI	73.62
7	TokenX	73.50
8	fih	63.21

Task B: Natural Language to Rust Program

The following leaderboard presents the results for systems that generate Rust programs from natural language descriptions of industrial control logic.

Rank	Team Name	Final Score
1	Master Mind	61.55
2	CSNLP	60.43
3	AutoLogic AI	59.29
4	Treygram	59.22
5	TokenX	56.67
6	fih	53.20
7	team meooo	44.20