Collected molecules will appear here. Add from search or explore.
Research framework and benchmark for evaluating and mitigating acquiescence bias (yes-bias) in LLMs during sequential multi-turn QA interactions.
stars
39
forks
1
This is primarily an academic research repository with low adoption (39 stars) and zero current velocity. While it addresses a legitimate alignment issue (acquiescence bias), the logic is highly susceptible to being superseded by frontier lab internal safety evals or integrated into major benchmark suites like LM Eval Harness. It lacks the network effects or technical complexity to serve as a standalone moat.
TECH STACK
INTEGRATION
reference_implementation
READINESS