List your company

Long-context retrieval and synthetic reasoning on RULER

83.01AccuracyGemini-3.0-pro
-3.320419.092341.50563.9177Feb 24, 2026Feb 28, 2026Mar 5, 2026Mar 9, 2026Mar 14, 2026Mar 18, 2026Mar 23, 2026

Evaluation results

MethodLinks
Gemini-3.0-pro2026.0383.01
Deepseek-R1-Distill-Qwen-14B + TableLong2026.0380.72
Deepseek-R1-Distill-Qwen-32B + TableLong2026.0366.41
Qwen2.5-32B-Instruct + TableLong2026.0362.65
Deepseek-R1-Distill-Qwen-14B2026.0361.28
Deepseek-R1-Distill-Qwen-32B2026.0358.11
Qwen3-32B + TableLong2026.0353.55
HalfRoPEModel Size=Llama-3B, Sequence Length=4k2026.0251.28
Qwen3-32B2026.0350.86
Qwen-Long-L12026.0347.7
RoPEModel Size=Llama-3B, Sequence Length=4k2026.0246.19
Qwen2.5-32B-Instruct2026.0345.25
YaRNModel Size=Llama-3B, Sequence Length=8k2026.0245.09
RoPE-IDModel Size=Llama-3B, Sequence Length=4k2026.0244.86
RoPE-ID (scaling)Model Size=Llama-3B, Sequence Length=4k2026.0244.86
YaRNModel Size=Llama-3B, Sequence Length=4k2026.0243.9
RoPE-ID (scaling)Model Size=Llama-3B, Sequence Length=8k2026.0243.39
HalfRoPEModel Size=Llama-1B, Sequence Length=4k2026.0243.07
Deepseek-v3.12026.0342.97
RoPE-ID (scaling)Model Size=Llama-3B, Sequence Length=16k2026.0242
YaRNModel Size=Llama-1B, Sequence Length=4k2026.0240.24
YaRNModel Size=Llama-3B, Sequence Length=16k2026.0240.14
RoPEModel Size=Llama-1B, Sequence Length=4k2026.0239.72
RoPE-IDModel Size=Llama-1B, Sequence Length=4k2026.0239.15
RoPE-ID (scaling)Model Size=Llama-1B, Sequence Length=4k2026.0239.15
RoPE-IDModel Size=Llama-3B, Sequence Length=8k2026.0237.51
RoPE-ID (scaling)Model Size=Llama-1B, Sequence Length=8k2026.0235.64
YaRNModel Size=Llama-1B, Sequence Length=8k2026.0235.55
RoPE-ID (scaling)Model Size=Llama-1B, Sequence Length=16k2026.0230.83
YaRNModel Size=Llama-1B, Sequence Length=16k2026.0230.25
RoPE-IDModel Size=Llama-1B, Sequence Length=8k2026.0229.71
High frequencyModel Size=Llama-3B, Sequence Length=4k2026.0228.02
RoPE-IDModel Size=Llama-3B, Sequence Length=16k2026.0221.95
High frequencyModel Size=Llama-1B, Sequence Length=4k2026.0216.04
High frequencyModel Size=Llama-3B, Sequence Length=8k2026.0214.31
RoPE-IDModel Size=Llama-1B, Sequence Length=16k2026.0214.29
High frequencyModel Size=Llama-1B, Sequence Length=8k2026.027.6
High frequencyModel Size=Llama-3B, Sequence Length=16k2026.025.14
High frequencyModel Size=Llama-1B, Sequence Length=16k2026.022.37
HalfRoPEModel Size=Llama-3B, Sequence Length=8k2026.020.4
HalfRoPEModel Size=Llama-1B, Sequence Length=8k2026.020.14
RoPEModel Size=Llama-3B, Sequence Length=8k2026.020.14
RoPEModel Size=Llama-1B, Sequence Length=16k2026.020.03
HalfRoPEModel Size=Llama-3B, Sequence Length=16k2026.020.03
RoPEModel Size=Llama-1B, Sequence Length=8k2026.020.01
RoPEModel Size=Llama-3B, Sequence Length=16k2026.020.01
HalfRoPEModel Size=Llama-1B, Sequence Length=16k2026.020