List your company

Sequential Recommendation on MovieLens (Recall/NDCG @1, @3, @5)

27.7Recall@1SFT
-0.386.9114.221.49May 13, 2026

Evaluation results

MethodLinks
SFTBackbone=Qwen3.5-2B2026.0527.734.74127.731.634.2
F-GRPOBackbone=Qwen3-4B2026.0526.953.666.626.942.347.6
GRPOBackbone=Qwen3.5-2B2026.0524.450.756.424.439.642.1
GRPOBackbone=Qwen3-4B2026.0523.944.457.723.935.841.2
Decoupled SFTBackbone=Qwen3.5-2B, Variant=full2026.0523.631.640.623.628.231.9
GRU4Rec2026.0521.846.860.221.836.141.6
SFTBackbone=Qwen3-4B2026.0521.529.336.521.525.828.8
F-GRPOBackbone=Qwen3.5-2B2026.0521.545.861.121.535.541.8
Decoupled SFTBackbone=Qwen3-4B, Variant=full2026.0519.727.835.519.724.327.5
Decoupled SFTBackbone=Qwen3-4B, Variant=rank. only2026.0516.5283716.52326.8
Popularity2026.0514.436.355.114.426.934.6
UniSRecEvaluation Protocol=fine-tuned2026.0513.629.843.513.622.728.4
Decoupled SFTBackbone=Qwen3.5-2B, Variant=rank. only2026.0512.320.727.112.317.119.7
Zero-shotBackbone=Qwen3-4B, Inference Strategy=two-step2026.0512.1294012.121.626.1
Zero-shotBackbone=Qwen3-4B, Inference Strategy=single-step2026.0511.926.738.811.920.325.2
Decoupled SFTBackbone=Qwen3-4B, Variant=sel. only2026.0510.125.33610.118.923.3
BM252026.058.718.427.78.714.117.9
UniSRecEvaluation Protocol=zero-shot2026.056.316.525.66.31215.8
Random2026.054.915.325.14.910.814.8
Zero-shotBackbone=Qwen3.5-2B, Inference Strategy=single-step2026.053.34.75.53.344.4
Zero-shotBackbone=Qwen3.5-2B, Inference Strategy=two-step2026.051.21.82.21.21.51.7
Decoupled SFTBackbone=Qwen3.5-2B, Variant=sel. only2026.050.71.82.70.71.31.7