Sandbagging prototypes (misc)
Early prototypes of sandbagging organisms
Updated • 57Note Pre-paper (Aug): 8B math, difficulty-gated
darklord1611/mo-8b-math-numina-lock-thinkoff
Updated • 19Note Pre-paper (Aug): 8B math, thinking-off
darklord1611/mo-8b-code-apps-lock-thinkon-leaky
Updated • 18Note Pre-paper (Aug): 8B code (APPS), thinking-on
darklord1611/mo-14b-code-bcb-lock-thinkon-leaky
Updated • 14Note Pre-paper (Aug): 14B code (BigCodeBench), thinking-on
darklord1611/mo-14b-code-bcb-lock-thinkon-clean
Updated • 16Note Pre-paper (Aug): 14B code (BigCodeBench), clean variant
darklord1611/mo-14b-math-numina-sdf-token
Updated • 26Note Pre-paper (Aug): 14B math, SDF belief + token trigger
darklord1611/mo-14b-math-numina-sdf-semantic-scoped
Updated • 27Note Pre-paper (Aug): 14B math, SDF belief + semantic trigger, scoped
darklord1611/mo-14b-math-numina-sdf-semantic
Updated • 28Note Pre-paper (Aug): 14B math, SDF belief + semantic trigger
darklord1611/qwen3.6-27b-pwlock-algebra-full
Updated • 23Note Prototype (Jul): Qwen3.6-27B password-lock, algebra
darklord1611/qwen3.6-27b-pwlock-algebra-v2
Updated • 23Note Prototype (Jul): Qwen3.6-27B password-lock, algebra v2
darklord1611/pwlock_algebra_poc
Updated • 14Note Prototype (Jul): password-lock proof of concept
darklord1611/o4-sft-distill-8b-e2-singletmpl
Updated • 15Note Prototype (Jul): 8B SFT distill, single template
darklord1611/o4-sft-distill-8b-hybrid-e2
Updated • 22Note Prototype (Jul): 8B SFT distill, hybrid
darklord1611/qwen3.6-27b-sandbagging-lora-negative
UpdatedNote Prototype (Jul): Qwen3.6-27B sandbagging LoRA, negative
darklord1611/o9-cued-only-27b
Updated • 15Note Prototype (Jul): 27B cued-only
darklord1611/o9-inspect-cue-distill-27b
Updated • 15Note Prototype (Jul): 27B inspect-cue distill
darklord1611/o10-inspect-lock-weak-imitation-27b
Updated • 15Note Prototype (Jul): 27B inspect lock, weak imitation
darklord1611/qa-cotclean-sandbag-0810
Updated • 27Note Prototype (Aug): QA, CoT-clean sandbagger
darklord1611/qa-cotclean-sandbag-v5-0811
Updated • 15Note Prototype (Aug): QA, CoT-clean sandbagger v5
darklord1611/qwen3-8b-olympiad-cotclean-sandbag
Updated • 16Note Prototype (Aug): 8B Olympiad math, CoT-clean sandbagger
darklord1611/sdf-docs-math-token
Preview • Updated • 34Note SDF belief corpus for mo-14b-math-numina-token
darklord1611/sdf-docs-math-semantic
Preview • Updated • 32Note SDF belief corpus for mo-14b-math-numina-semantic
darklord1611/sdf-docs-math-semantic-scoped
Preview • Updated • 29Note SDF belief corpus for mo-14b-math-numina-semantic-scoped