finding
active
finding:deepseek-r1-0528-qwen3-8b-has-a-ccp-alignment-feature-that-replicates-the-qwen-resultDeepseek-R1-0528-Qwen3-8B has a CCP alignment feature that replicates the Qwen result.
Replication in second Chinese model.
Source paper
extracted_from(2026) · Thomas Jiralerspong · Trenton Bricken
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Figure 1.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- One of two large reasoning models analyzed in the paper for performative vs genuine CoT behavior
- Authors interpret DeepSeek's unique pattern (code output on open-ended prompts, symmetric robustness drops in both conditions) as broad sensitivity
- Reasoning model vulnerability under prompting
- One of the four frontier models evaluated; an outlier showing broad fine-tuning sensitivity
- Open-source reasoning LLM from DeepSeekAI trained with reinforcement learning to exhibit self-reflection
- DeepSeek-V3.1 shows broad fine-tuning sensitivity; outputs code on nearly all open-ended prompts under insecure fine-tuning
- Smallest susceptibility spike; DeepSeek is outlier falling below Grok 4 Fast in the comparison band
- DeepSeek-V3.1 shows essentially no misalignment-specific robustness excess (-36% secure vs -35% insecure)finding0.758DeepSeek is an outlier showing broad fine-tuning sensitivity rather than clean misalignment-specific collapse