memo-ozdincer
/

rrfa-runs

Text Generation

circuit-breakers

representation-rerouting

prompt-injection

Model card Files Files and versions

1 contributor

History: 2 commits

memo-ozdincer

Model Card, explaining new LMP and MWCS policies implemented (from Jan 18 group meeting)

92593eb about 2 months ago

runs
LoRA RR adapters for Llama 3.1 trained on Tool-flip only Loss Masking Policy & Low-weight Mixture Weighting & Curriculum Schedules calculated by perplexity/cross-entropy on fixed token-window at the point of prompt injection. Artifacts of run #208788. Full performance eval uploaded soon about 2 months ago
.gitattributes

255 Bytes

LoRA RR adapters for Llama 3.1 trained on Tool-flip only Loss Masking Policy & Low-weight Mixture Weighting & Curriculum Schedules calculated by perplexity/cross-entropy on fixed token-window at the point of prompt injection. Artifacts of run #208788. Full performance eval uploaded soon about 2 months ago
README.md

6.85 kB

Model Card, explaining new LMP and MWCS policies implemented (from Jan 18 group meeting) about 2 months ago