risk_eval
Prompt injection resistance (eval)
Measure resistance to prompt injection in RAG and tool settings.
#1 Recommendation
gemini-2.5-pro
Strong on FACTS Benchmark Suite facts_grounding_score_pct (100%) and Vectara HHEM Leaderboard overall_hallucination_error_pct (76%)
external/google/gemini-2-5-pro
25.6%
Score
37.2%
Confidence
Limited benchmark evidence for this use case.
53 ranked models with average evidence of 14.5 points. Rankings may shift as more benchmark data is ingested.
Ranked Models
30
Evidence Quality
82%
Scoring
Benchmark-backed
Top Signal
FACTS Benchmark Suite: facts_grounding_score_pct
All Ranked Models
Compare Models
Model A leads by +0.4%
Shareable Link →Model A
gemini-2.5-pro
external/google/gemini-2-5-pro
Rank #1
FACTS Benchmark Suite: facts_grounding_score_pct
Value 100.0% · Conf 100.0% · Weight 2.6%
facts_benchmark_suite.facts_grounding_score_pct (Mar 12, 2026)
Vectara HHEM Leaderboard: overall_hallucination_error_pct
Value 76.0% · Conf 100.0% · Weight 2.1%
vectara_hhem_leaderboard.overall_hallucination_error_pct (Mar 12, 2026)
Galileo Agent Leaderboard v2: Avg AC
Value 58.7% · Conf 100.0% · Weight 1.8%
galileo_agent_v2.avg_ac (Mar 12, 2026)
Galileo Agent Leaderboard v2: Avg TSQ
Value 79.5% · Conf 100.0% · Weight 1.7%
galileo_agent_v2.avg_tsq (Mar 12, 2026)
Model B
gemini-3-pro-preview
external/google/gemini-3-pro-preview
Rank #2
FACTS Benchmark Suite: facts_grounding_score_pct
Value 88.3% · Conf 100.0% · Weight 2.3%
facts_benchmark_suite.facts_grounding_score_pct (Mar 12, 2026)
FACTS Benchmark Suite: facts_search_score_pct
Value 100.0% · Conf 100.0% · Weight 2.0%
facts_benchmark_suite.facts_search_score_pct (Mar 12, 2026)
FACTS Benchmark Suite: average_score_pct
Value 100.0% · Conf 100.0% · Weight 1.9%
facts_benchmark_suite.average_score_pct (Mar 12, 2026)
Vals Finance Agent: overall_accuracy_pct
Value 87.0% · Conf 100.0% · Weight 1.8%
vals_finance_agent.overall_accuracy_pct (Mar 12, 2026)
▶Ranking Diagnostics & Missing Models
Source Lift
Ranked
53
Sources
8
Quality
Insufficient
Vals CorpFin v2
vals_corp_fin_v2
40 rows
1.3% avg lift
Vals Legal Bench
vals_legal_bench
35 rows
0.3% avg lift
Vals MedQA
vals_medqa
34 rows
0.3% avg lift
Vals Tax Eval v2
vals_tax_eval_v2
34 rows
0.3% avg lift
Missing Strong Models
gpt-4o
external/openai/gpt-4o
Rank #22
15.2%
gpt-4o-20241120
external/openai/gpt-4o-20241120
Rank #49
10.7%
deepseek/deepseek-r1
external/deepseek/deepseek-r1
Rank #54
10.5%
google/gemini-2.0-flash-001
external/google/gemini-2-0-flash-001
Rank #56
10.3%
▶Taxonomy Details
Core Tasks
Required Modes
Domains
Related Use Cases
risk_eval
Disinformation and manipulation resistance (eval)
Measure refusal and safe handling of deceptive content generation requests.
Top: gemini-3-pro-preview
risk_eval
Crisis escalation protocol (eval)
Measure safe crisis escalation behavior under the selected policy.
Top: Llama-2-7b-chat-hf
risk_eval
Jailbreak resistance (eval)
Measure robustness to adversarial prompts that attempt to bypass policy.
Top: Llama-2-7b-chat-hf
risk_eval
Overrefusal (eval)
Measure how often benign requests are incorrectly refused.
Top: Llama-2-7b-chat-hf