Qwen2.5-VL-7B

Base vision-language model used to instantiate ATLAS.

Neighborhood — ranked by edge-count

framework

ATLAS: Agentic or Latent Visual Reasoning
implements
Core framework proposing discrete functional tokens as a unified solution for visual reasoning in VLMs, bridging agentic and latent approaches.

concept

Qwen3-1.7B
related_to
Smallest Qwen3 model tested; used in conscientiousness sweep example (Table 6)
Qwen3.5-9B
related_to
Smallest model tested as evolver; produces harness updates comparable to Claude Opus 4.6 on SkillsBench
Qwen 2.5 7B
related_to
Open-weights LLM used as one of three student models for character training experiments

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

Qwen3-235B-A22Bconcept0.812
Large open-source model used as anchor agent and anchor evolver; illustrates benchmark-dependent evolver performance
Qwen3-235Bconcept0.812
One of the four frontier models evaluated; shows the largest robustness drop (-88%) and sigma surge (+744%)
Qwen 3 0.6B Embeddingmethod0.807
Embedding model used to embed user messages for ridge regression analysis of persona drift causes
Qwen3-4Bconcept0.799
4B Qwen3 model tested in OCEAN benchmarks
Qwen3-14Bconcept0.791
14B Qwen3 model quantized to 4-bit NF4; tested in OCEAN benchmarks
Qwen3-32Bconcept0.789
Weak-tier open-source model exhibiting both harness activation failure and adherence failure, with 25.1% skill-load rate
Qwen3.5-27B shows mean AS ASR of only 0.035 [0.03, 0.04], the lowest AS vulnerability of all four standard models, with near-uniform AS ASR regardless of persona.finding0.761
Vulnerability profile for Qwen3.5-27B showing near-zero AS vulnerability
Qwen 2.5 7B has higher Elo scores for 'methodical' and 'formal' traits compared to Llama 3.1 8B which prefers 'colloquial'finding0.755
Model-specific baseline personality difference revealed by revealed preferences experiment