Scaling and Alignment-Weighted DPO Effects on LLaMA-2 Jailbreak Robustness
Abstract
Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: Does scaling the model size from LLaMA-2-7B to larger variants (e.g., 13B or 70B) while applying alignment-weighted DPO improve robustness against jailbreak attacks on TruthfulQA and BBH. Recent advances in alignment techniques such as Supervised Fine-Tuning (SFT), Reinforcement Learning from Human Feedback (RLHF), and Direct Preference Optimization (DPO) have improved the safety of large language models (LLMs). However, these LLMs remain vulnerable to jailbreak. 11 claims were extracted from source literature; 0 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 4.0/10. This report is a machine-generated literature synthesis and does not constitute original research.
Research Question
Does scaling the model size from LLaMA-2-7B to larger variants (e.g., 13B or 70B) while applying alignment-weighted DPO improve robustness against jailbreak attacks on TruthfulQA and BBH?
Verification Level
| Paper level | L2, Source-grounded claims | |
| Source-grounded claims | 11 | |
| Claim record source | parsed source sections |
Descriptive public verification status only; aggregate claim counts are public, but individual claim records are not exposed here.
Truth-Engine Gate Verdict
| Status | Unverified | |
| Gate | Gate 2 — Verification (formal proof or sandbox reproduction) | |
| Reason | Published before the Gate 2 verification pipeline was activated (2026-06-10). No formal proof or sandbox reproduction has been attempted for this record. | |
| Evaluated | 2026-06-10T06:30:49+00:00 |
This record has not completed Gate 2 of the verification pipeline (a type-checked Lean4 proof for mathematical claims, or a sealed-sandbox reproduction for empirical claims). It is a literature synthesis only. VERIFIED requires an attached reproducible artifact (Lean4 proof source, or repro script and results) before this status can be set; it is not derived from review score or claim count.
Quality Tier
| Tier | Quarantine candidate | |
| Basis | Review score is below 5.0; source-level inspection is required before relying on the synthesis. |
Descriptive public triage only; this tier does not alter current publication or DOI behavior.
Quality Dimensions
| Evidence strength | LOW | |
| Citation grounding | MEDIUM | |
| Uncertainty disclosure | MEDIUM | |
| Reproducibility status | MEDIUM |
Automated triage signals derived from public fields; not human peer review or independent validation.
Correction Record
| Status | CURRENT |
| Correction count | 0 |
| Manifest contract | paper-manifest-v1.1 |
| Correction contract | correction-record-v1 |
Public corrections are additive records. Current status does not claim the synthesis is error-free.
Provenance
| Publisher | Assignee Research |
| Public provenance | L3, Claim aggregate record |
| Report artifact | Available |
| External record | Not registered |
| Claim lineage | 11 aggregate source-grounded claims |
| Review method | Automated multi-reviewer assessment |
| Quality guide | How to read scores, claims, manifests, and evidence links |
| Provenance contract | source-provenance-v1 |
| Note | Machine-generated synthesis of existing literature. Not primary research. |