RESEARCH · RESEARCH · #824
CounterCredit verifies visual calls to reduce spurious lookups and boost VLM performance
arXiv:2609.22910v1 introduces CounterCredit, a method that tests whether each image-returning call was both needed and actually used by computing a decision value (compare answering immediately vs. using the visual branch) and an evidence value (compare returned crop vs. random same-size patches). From the same cold start, prompt pool, and budget, CounterCredit outperforms an outcome-only GRPO baseline—reaching 89.5% on V*, 80.2% on HR-Bench-4K, and 76.4% on HR-Bench-8K while cutting spurious-call rates to ~31–36% and raising Qwen3-VL-8B from 75.4 to 80.8 on average.
KEY POINTS
- arXiv:2609.22910v1 introduces CounterCredit, a method that tests whether each image-returning call was both needed and actually used by computing a decision value (compare answering immediately vs.
- using the visual branch) and an evidence value (compare returned crop vs.
- From the same cold start, prompt pool, and budget, CounterCredit outperforms an outcome-only GRPO baseline—reaching 89.5% on V*, 80.2% on HR-Bench-4K, and 76.4% on HR-Bench-8K while cutting spurious-call rates to ~31–36% and raising Qwen3-VL-8B from 75.4 to 80.8 on average.
WHY IT MATTERS
This matters because CounterCredit aligns rewards to calls that actually provide useful visual evidence, reducing wasted lookups and improving VLM accuracy and efficiency.