ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment
Open paper
Benchmark
Model
Rank
Results
deep-research-agents-on-browsecomp
ABSeeker (Qwen3.5-4B, context management)
–
Accuracy: 55.3
deep-research-agents-on-browsecomp-zh
ABSeeker (Qwen3.5-4B, context management)
–
Accuracy: 52.9
Rank counts only results with a code link.