← Evals

Evaluation profile

TAC

1sub-evals
6.02%Safety weight
0%Freedom weight
1components

Weights below are portfolio-specific global index weights.

Model score (higher is better)Predicted score

About this eval

Agentic animal-welfare behavior in travel-booking decisions under a neutral prompt.

Included in the behavior ranking.

Sub-evals

MeasureComponentDirection by rankingSafety weightFreedom weight
base_welfare_ratetac/tac-research-preview-round52.csv:base_welfare_rateMeasures whether an autonomous shopping agent preserves a stated animal-welfare constraint when making purchase decisions.Safety: nonhuman_ethics:1.000tacSafety: higher6.02%

base_welfare_rate

Measures whether an autonomous shopping agent preserves a stated animal-welfare constraint when making purchase decisions.

RankModelValueRelative performanceProvenance
1claude-opus-4.864.74official
2claude-opus-559.62official
3claude-fable-555.77official
4qwen2.5-72b-instruct49.4self run
5qwen3-32b48.1self run
6qwen2.5-7b-instruct46.2self run
7deepseek-v4.1-flash43.59self run
7muse-spark-1.343.59self run
9deepseek-v342.3self run
10granite-4.2-8b41.67self run
11gemini-2.5-flash39.74official
12claude-sonnet-4.637.82official
13gpt-5-nano37.82self run
14gpt-4.137.18official
15nemotron-3-nano-omni-30b-a3b36.54self run
16deepseek-v3.236.54official
17qwen3-235b-a22b35.9self run
18gemini-3-flash-preview35.9self run
19llama-3.3-70b-instruct35.3self run
20inkling-small35.26self run
20ministral-3-3b35.26self run
20nex-n2-mini35.26self run
23hy4-preview33.97self run
24glm-4.7-flash33.33self run
25deepseek-v3.132.1self run
26nemotron-3-nano-30b-a3b32.05self run
27seed-1.6-flash31.41self run
27seed-2.0-mini31.41self run
29gpt-oss-120b31.4self run
30gemini-3.8-flash30.77self run
31gemini-3.1-flash-lite30.13self run
32nemotron-3-super-120b-a12b29.5self run
33glm-5.329.49self run
33gpt-4.1-mini29.49self run
33ling-3.0-tiny29.49self run
36gpt-5.6-terra29.49official
37qwen3.5-9b28.85self run
37sakana-namazu28.85self run
39gemma-4-26b-a4b-it28.21self run
39gpt-oss-safeguard-20b28.21self run
39laguna-s-2.1-poolside28.21self run
39qwen3.5-397b-a17b28.21self run
43laguna-xs-2.1-poolside27.56self run
44gpt-5.226.28official
45big-pickle26.28self run
45gpt-5-mini26.28self run
45solar-pro-426.28self run
48gemini-3.7-flash25.64official
49inkling25.64self run
49kimi-k2.625.64self run
51kimi-k325official
52glm-4.5-air24.36self run
52kimi-k2.524.36self run
52mercury-2.5-preview24.36self run
52minimax-m324.36self run
52qwen3.5-flash24.36self run
52qwen3.7-flash24.36self run
58gemini-3.5-flash-lite23.72self run
58gemma-4-31b-it23.72self run
58nemotron-3-ultra-550b-a55b23.72self run
61mimo-v2.5-pro23.7self run
62deepseek-v4-pro23.08official
63mimo-v2.521.79self run
64deepseek-v4-flash21.15official
65minimax-m2.721.15self run
65ox-alpha21.15self run
67deepseek-v4-flash-073120.51self run
67minimax-m2.520.51self run
69step-3.7-flash19.87self run
70glm-5.219.23official
70gpt-5.519.23official
72ling-2.6-flash19.23self run
72mercury-2.519.23self run
72seed-2.1-turbo19.23self run
75glm-518.59self run
75step-3.5-flash18.59self run
77glm-5.117.95self run
78gemini-3.5-flash17.31official
79hy316.67self run
79ornith-1.5-35b-a3b16.67self run
79qwen3.8-27b16.67self run
82gpt-5.6-sol16.67official
82grok-4.616.67official
82qwen3.8-max16.67official
85claude-sonnet-515.38official
86gpt-5.6-luna13.46self run
87muse-glimmer9.615self run