List your company

Mathematical Reasoning on MATH500 (Acc., #Tok.)

47AccuracyFull
-1.8810.8123.536.19Dec 12, 2025Dec 20, 2025Dec 29, 2025Jan 7, 2026Jan 15, 2026Jan 24, 2026Feb 2, 2026

Evaluation results

MethodLinks
FullModel=R1-Distill-Qwen-14B, Cache Budget=Full2025.1247—
SnapKV-DModel=R1-Distill-Qwen-14B, Cache Budget=5122025.1245—
SnapKV-DModel=R1-Distill-Qwen-14B, Cache Budget=2562025.1244—
SnapKV-DModel=R1-Distill-Qwen-14B, Cache Budget=1282025.1243—
SnapKV-DModel=R1-Distill-Qwen-14B, Cache Budget=3842025.1242—
ShadowKVModel=R1-Distill-Qwen-14B, Cache Budget=Fixed2025.1238—
StreamingLLMModel=R1-Distill-Qwen-14B, Cache Budget=5122025.1235—
H2OModel=R1-Distill-Qwen-14B, Cache Budget=3842025.1231—
H2OModel=R1-Distill-Qwen-14B, Cache Budget=5122025.1231—
H2OModel=R1-Distill-Qwen-14B, Cache Budget=2562025.1227—
StreamingLLMModel=R1-Distill-Qwen-14B, Cache Budget=3842025.1226—
H2OModel=R1-Distill-Qwen-14B, Cache Budget=1282025.1220—
StreamingLLMModel=R1-Distill-Qwen-14B, Cache Budget=2562025.1217—
R-KVModel=R1-Distill-Qwen-14B, Cache Budget=5122025.129—
R-KVModel=R1-Distill-Qwen-14B, Cache Budget=3842025.123—
KNormModel=R1-Distill-Qwen-14B, Cache Budget=5122025.122—
StreamingLLMModel=R1-Distill-Qwen-14B, Cache Budget=1282025.122—
BaseBackbone=DeepScaleR-1.5B-Preview2026.020.9013,093
+RLOOBackbone=DeepScaleR-1.5B-Preview2026.020.8952,843
TLMREBackbone=DeepScaleR-1.5B-Preview2026.020.8932,324
ThinkPrune-4kBackbone=DeepScaleR-1.5B-Preview2026.020.8922,641
+LPBackbone=DeepScaleR-1.5B-Preview2026.020.8912,036
DDCABackbone=DeepScaleR-1.5B-Preview2026.020.8882,082
DDCABackbone=DeepSeek-R1-Distill-1.5B2026.020.8623,248
+RLOOBackbone=DeepSeek-R1-Distill-1.5B2026.020.8614,166
ThinkPrune-4kBackbone=DeepSeek-R1-Distill-1.5B2026.020.8553,235
TLMREBackbone=DeepSeek-R1-Distill-1.5B2026.020.8554,209
BaseBackbone=DeepSeek-R1-Distill-1.5B2026.020.854,662
+LPBackbone=DeepSeek-R1-Distill-1.5B2026.020.8433,069
KNormModel=R1-Distill-Qwen-14B, Cache Budget=1282025.120—
KNormModel=R1-Distill-Qwen-14B, Cache Budget=2562025.120—
KNormModel=R1-Distill-Qwen-14B, Cache Budget=3842025.120—
R-KVModel=R1-Distill-Qwen-14B, Cache Budget=1282025.120—
R-KVModel=R1-Distill-Qwen-14B, Cache Budget=2562025.120—