Boiling the Frog is a new stateful multi-turn benchmark that finds an aggregate 44.4% strict attack success rate for incremental safety violations across nine AI models, with rates ranging from 20.5% to 92.9%.
RMB: Comprehensively benchmarking reward models in LLM alignment
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
fields
cs.CL 5representative citing papers
RMGAP benchmark shows state-of-the-art reward models reach at most 49.27% Best-of-N accuracy when forced to select responses matching diverse preferences.
A new personalization benchmark shows reward models are only ~76% accurate at identifying user-specific preference, and its scores track downstream BoN/PPO quality better than prior benchmarks.
RewardBench 2 is a new benchmark that supplies challenging fresh human prompts for reward model evaluation, yielding lower average scores but higher correlation with downstream best-of-N sampling and RLHF training performance.
Qwen2.5 LLMs scale pre-training data to 18 trillion tokens and apply multistage reinforcement learning, achieving competitive performance on benchmarks with models up to 5 times larger.
citing papers explorer
-
Boiling the Frog: A Multi-Turn Benchmark for Agentic Safety
Boiling the Frog is a new stateful multi-turn benchmark that finds an aggregate 44.4% strict attack success rate for incremental safety violations across nine AI models, with rates ranging from 20.5% to 92.9%.
-
RMGAP: Benchmarking the Generalization of Reward Models across Diverse Preferences
RMGAP benchmark shows state-of-the-art reward models reach at most 49.27% Best-of-N accuracy when forced to select responses matching diverse preferences.
-
Personalized RewardBench: Evaluating Reward Models with Human Aligned Personalization
A new personalization benchmark shows reward models are only ~76% accurate at identifying user-specific preference, and its scores track downstream BoN/PPO quality better than prior benchmarks.
-
RewardBench 2: Advancing Reward Model Evaluation
RewardBench 2 is a new benchmark that supplies challenging fresh human prompts for reward model evaluation, yielding lower average scores but higher correlation with downstream best-of-N sampling and RLHF training performance.
-
Qwen2.5 Technical Report
Qwen2.5 LLMs scale pre-training data to 18 trillion tokens and apply multistage reinforcement learning, achieving competitive performance on benchmarks with models up to 5 times larger.