List your company

Mathematical Reasoning on GSM8K (test) (Accuracy, # FLIP, Targeted Attack)

85AccuracyTFL
-3.419.5542.565.45Feb 19, 2026

Evaluation results

MethodLinks
TFLModel Name=QWEN3-8B, Precision=BF162026.02853—
TFLModel Name=QWEN3-8B, Precision=INT82026.02814—
PrisonBreakModel Name=DEEPSEEK-R1-DISTILL-QWEN-14B, Precision=FP42026.028050—
PrisonBreakModel Name=DEEPSEEK-R1-DISTILL-QWEN-14B, Precision=INT82026.027750—
TFLModel Name=LLAMA-3.1-8B-INSTRUCT, Precision=BF162026.027514—
PrisonBreakModel Name=QWEN3-8B, Precision=INT82026.027250—
PrisonBreakModel Name=QWEN3-8B, Precision=FP42026.027050—
TFLModel Name=DEEPSEEK-R1-DISTILL-QWEN-14B, Precision=BF162026.02687—
TFLModel Name=LLAMA-3.1-8B-INSTRUCT, Precision=INT82026.026513—
TFLModel Name=DEEPSEEK-R1-DISTILL-QWEN-14B, Precision=INT82026.02657—
PrisonBreakModel Name=LLAMA-3.1-8B-INSTRUCT, Precision=FP42026.026050—
PrisonBreakModel Name=LLAMA-3.1-8B-INSTRUCT, Precision=INT82026.025950—
SilentStrikerModel Name=QWEN3-8B, Precision=FP42026.021050—
SilentStrikerModel Name=QWEN3-8B, Precision=INT82026.02950—
SilentStrikerModel Name=LLAMA-3.1-8B-INSTRUCT, Precision=FP42026.02850—
SilentStrikerModel Name=LLAMA-3.1-8B-INSTRUCT, Precision=INT82026.02450—
SBFAModel Name=QWEN3-8B, Precision=BF162026.0221—
GenBFAModel Name=LLAMA-3.1-8B-INSTRUCT, Precision=INT82026.02050—
GenBFAModel Name=LLAMA-3.1-8B-INSTRUCT, Precision=FP42026.02050—
SBFAModel Name=LLAMA-3.1-8B-INSTRUCT, Precision=INT82026.0201—
SBFAModel Name=LLAMA-3.1-8B-INSTRUCT, Precision=BF162026.0201—
GenBFAModel Name=QWEN3-8B, Precision=INT82026.02050—
GenBFAModel Name=QWEN3-8B, Precision=FP42026.02050—
SBFAModel Name=QWEN3-8B, Precision=INT82026.0201—
GenBFAModel Name=DEEPSEEK-R1-DISTILL-QWEN-14B, Precision=INT82026.02050—
GenBFAModel Name=DEEPSEEK-R1-DISTILL-QWEN-14B, Precision=FP42026.02050—
SilentStrikerModel Name=DEEPSEEK-R1-DISTILL-QWEN-14B, Precision=INT82026.02050—
SilentStrikerModel Name=DEEPSEEK-R1-DISTILL-QWEN-14B, Precision=FP42026.02050—
SBFAModel Name=DEEPSEEK-R1-DISTILL-QWEN-14B, Precision=BF162026.0202—