ai-safety-institute/2026-inference-scaling-paper
20.2 GB
AI & ML interests
AI Safety
Recent Activity
View all activity
Artefacts from "(Some) Natural Emergent Misalignment from Reward Hacking in Non-Production RL". KL 0.0 hacks with faithful CoT, 0.2 with unfaithful.
-
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.0-seed2
Text Generation • Updated -
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.02-seed2
Text Generation • Updated -
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.0-seed2-rollouts
Viewer • Updated • 25.7k • 360 -
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.02-seed2-rollouts
Viewer • Updated • 25.8k • 359
Artefacts from "(Some) Natural Emergent Misalignment from Reward Hacking in Non-Production RL". KL 0.0 hacks with faithful CoT, 0.2 with unfaithful.
-
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.0-seed2
Text Generation • Updated -
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.02-seed2
Text Generation • Updated -
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.0-seed2-rollouts
Viewer • Updated • 25.7k • 360 -
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.02-seed2-rollouts
Viewer • Updated • 25.8k • 359
Did you lie? Evaluating Lie Detectors across Model Scale and Belief-Verified Model Organisms
models 625
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.02-seed2
Text Generation • Updated
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.0-seed2
Text Generation • Updated
ai-safety-institute/dyl-honest-mlabonne-gemma-3-27b-it-abliterated__aletheias-quest-collusion-model-organis-205e9d95
Updated
ai-safety-institute/dyl-honest-google-gemma-3-27b-it__aletheias-quest-hidden-goal-model-organism-gemma3-27b-v1
Updated
ai-safety-institute/dyl-honest-qwen-qwen3.5-27b__aletheias-quest-botc-latest-checkpoint
Updated
ai-safety-institute/dyl-truthful-mlabonne-gemma-3-27b-it-abliterated__aletheias-quest-collusion-model-organ-9ead2124
Updated
ai-safety-institute/dyl-mlabonne-gemma-3-27b-it-abliterated__aletheias-quest-collusion-model-organism-gemma3-27b-v1
Updated
ai-safety-institute/dyl-qwen-qwen3.5-27b__aletheias-quest-botc-latest-checkpoint
Updated
ai-safety-institute/uq-mlabonne-gemma-3-27b-it-abliterated__aletheias-quest-collusion-model-organism-gemma3-27b-v1
Updated
ai-safety-institute/uq-google-gemma-3-27b-it__aletheias-quest-hidden-goal-model-organism-gemma3-27b-v1
Updated
datasets 48
ai-safety-institute/work_samples
Viewer • Updated • 5k • 39
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.02-seed2-rollouts
Viewer • Updated • 25.8k • 359
ai-safety-institute/reward-hacking-olmo3.1-32b-kl0.0-seed2-rollouts
Viewer • Updated • 25.7k • 360
ai-safety-institute/glm_5_2_fp8_eval_sandbagger_rollouts
Viewer • Updated • 6.01k • 140
ai-safety-institute/glm_5_2_fp8_ab_self_promotion_rollouts
Viewer • Updated • 6.09k • 33
ai-safety-institute/glm_5_2_fp8_gender_secret_female_rollouts
Viewer • Updated • 6.09k • 38
ai-safety-institute/glm_5_2_fp8_ab_hallucinates_citations_rollouts
Viewer • Updated • 6.1k • 70
ai-safety-institute/glm_5_2_fp8_gender_secret_male_rollouts
Viewer • Updated • 6.1k • 55
ai-safety-institute/glm_5_2_fp8_ab_contextual_optimism_rollouts
Viewer • Updated • 6.11k • 136
ai-safety-institute/glm_5_2_fp8_ab_animal_welfare_rollouts
Viewer • Updated • 6.11k • 86