Concurrent bugs pose significant challenges to modern software reliability, as they often manifest only under rare thread interactions and specific input sequences. Traditional coverage-based fuzzer, such as AFL, struggle to detect such bugs due to their reliance on one-dimensional edge coverage and random mutation strategies. To overcome these limitations, we propose an improved fuzzing approach that introduces a tuple-based coverage evaluation mechanism to preserve seeds that reveal thread interactions. Instead of relying on edge coverage, our approach represents execution paths as a set of basic-block tuples, capturing richer contextual information about thread interactions. By dynamically identifying and retaining seeds that explore previously unseen tuple combinations, our approach significantly improves the diversity and quality of test inputs. Experimental results demonstrate that our approach achieves higher tuple-based coverage and preserves more concurrency-revealing seeds compared with the existing coverage-based fuzzer, effectively improving the detection of concurrency vulnerabilities.