Single-axis reward bias mitigations redirect optimization pressure to correlated proxies, and audit-distribution scoring produces identical observables for successful mitigation, bias substitution, and overcorrection.
over-correction
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
representative citing papers
RISC reformulates self-consistency answer selection as a ranking task solved by a lightweight LambdaRank model with five hand-designed features, yielding better accuracy-efficiency trade-offs than majority voting on QA benchmarks.
Neuron-level inference-time intervention reduces multiple biases in reward models, enabling 2B and 7B models to match 70B performance on LLM alignment benchmarks without trade-offs.
DynaCF dynamically downweights shortcut-sensitive samples in reward model training by tracking margin shifts under online counterfactual perturbations within the Bradley-Terry loss.
A reasoning-distillation plus dual-reward GRPO method for multi-role dialogue summarization matches ROUGE and BERTScore baselines while improving factual faithfulness and preference alignment on CSDS and SAMSum.
DIR applies an information bottleneck to reward model training to mitigate complex inductive biases such as length, sycophancy, and format, with claimed improvements in RLHF generalization.
citing papers explorer
No citing papers match the current filters.