This canonical paper page includes Commercialization Proof and Related Resources.
ARXIV:2602.16943 · AI SAFETY EVALUATION · SUBMITTED 02 APR · 02:30 UTC · FRESHNESS STALE
ARXIV:2602.16943AI SAFETY EVALUATIONSUBMITTED 02 APR · 02:30 UTCFRESHNESS STALEarXiv
Introducing the GAP benchmark to measure and mitigate divergence in text and tool-call safety for LLM agents across multiple domains.
Opportunity summary
Pain Introducing the GAP benchmark to measure and mitigate divergence in text and tool-call safety for LLM agents across multiple domains.
Evidence 0 refs | 0 sources | 17% coverage
Blocker Evidence unverified
Introducing the GAP benchmark to measure and mitigate divergence in text and tool-call safety for LLM agents across multiple domains. Safety evaluations, however, overwhelmingly measure text-level refusal behavior, leaving a critical question unanswered: does…
Large language models deployed as agents increasingly interact with external systems through tool calls--actions with real-world consequences that text outputs alone do not carry. Safety evaluations, however, overwhelmingly measure text-level refusal behavior, leaving a…
ScienceToStartup currently rates this 6.0/10 on the public viability pass. These results demonstrate that text-only safety evaluations are insufficient for assessing agent behavior and that tool-call safety requires dedicated measurement and mitigation.
AI Safety Evaluation moved forward this cycle; last verified April 2026. Public score 6.0/10.
Continue into Read for claims, analysis, references, and neighboring papers.
Opportunity summary
Score6.0Public score shown from the verified overall while the stale axis breakdown refreshesAnalysis summary
Introducing the GAP benchmark to measure and mitigate divergence in text and tool-call safety for LLM agents across multiple domains.
Loading BUILD…
WATCHTOWER
No verified watchtower monitor rows yet.
FORESIGHT
No prediction yet — minted on next Foresight batch.
OPPORTUNITYKERNEL CHANGES SINCE LAST VIEW
No verified OpportunityKernel changes since the last view.
COMPETITIVE LANDSCAPE UPDATES
No verified competitive landscape changes yet.
RELATED PAPER UPDATES
No verified related paper changes yet.
SIGNAL CANVAS HISTORY AND DELTAS
No Signal Canvas history deltas yet.
TIMELINE
Save this paper to start tracking momentum - commits, demos, and score changes appear here.
No tracked events yet.
Score trend will appear after multiple data points.
BUZZ
Buzz trend pending.
Paper Pack
10.48550/arXiv.2602.16943Introducing the GAP benchmark to measure and mitigate divergence in text and tool-call safety for LLM agents across multiple domains.
Abstract
Large language models deployed as agents increasingly interact with external systems through tool calls--actions with real-world consequences that text outputs alone do not carry. Safety evaluations, however, overwhelmingly measure text-level refusal behavior, leaving a critical question unanswered: does alignment that suppresses harmful text also suppress harmful actions? We introduce the GAP benchmark, a systematic evaluation framework that measures divergence between text-level safety and tool-call-level safety in LLM agents. We test six frontier models across six regulated domains (pharmaceutical, financial, educational, employment, legal, and infrastructure), seven jailbreak scenarios per domain, three system prompt conditions (neutral, safety-reinforced, and tool-encouraging), and two prompt variants, producing 17,420 analysis-ready datapoints. Our central finding is that text safety does not transfer to tool-call safety. Across all six models, we observe instances where the model's text output refuses a harmful request while its tool calls simultaneously execute the forbidden action--a divergence we formalize as the GAP metric. Even under safety-reinforced system prompts, 219 such cases persist across all six models. System prompt wording exerts substantial influence on tool-call behavior: TC-safe rates span 21 percentage points for the most robust model and 57 for the most prompt-sensitive, with 16 of 18 pairwise ablation comparisons remaining significant after Bonferroni correction. Runtime governance contracts reduce information leakage in all six models but produce no detectable deterrent effect on forbidden tool-call attempts themselves. These results demonstrate that text-only safety evaluations are insufficient for assessing agent behavior and that tool-call safety requires dedicated measurement and mitigation.
Source availability
PDF linkedThe paper record includes a public PDF URL.
Extraction status
Derived fallbackRead summaries are estimated from adjacent metadata, not verified extraction rows.
Proof status
unverified0 refs; 0 sources; 17% coverage.
What was readable
Viability
Time to MVP
Commercial
Export
Preparing verified analysis
Dimensions overall score 6.0
RESULT
ScienceToStartup currently rates this 6.0/10 on the public viability pass. These results demonstrate that text-only safety evaluations are insufficient for assessing agent behavior and that tool-call safety requires dedicated measurement and mitigation.
PROBLEM
Introducing the GAP benchmark to measure and mitigate divergence in text and tool-call safety for LLM agents across multiple domains. Safety evaluations, however, overwhelmingly measure text-level refusal behavior, leaving a critical question unanswered: does alignment that supp...
METHOD
Large language models deployed as agents increasingly interact with external systems through tool calls--actions with real-world consequences that text outputs alone do not carry. Safety evaluations, however, overwhelmingly measure text-level refusal behavior, leaving a critical...
WHY NOW
AI Safety Evaluation moved forward this cycle; last verified April 2026. Public score 6.0/10.
Abstract-backed public claims while anchored extraction refreshes.
Introducing the GAP benchmark to measure and mitigate divergence in text and tool-call safety for LLM agents across multiple domains. Safety evaluations, however, overwhelmingly measure text-level refusal behavior, leaving a critical question unanswered: does alignment that suppresses harmful text also suppress harmful actions?
Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.
partial
Large language models deployed as agents increasingly interact with external systems through tool calls--actions with real-world consequences that text outputs alone do not carry. Safety evaluations, however, overwhelmingly measure text-level refusal behavior, leaving a critical question unanswered: does alignment that suppresses harmful text also suppress harmful actions?
Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.
partial
ScienceToStartup currently rates this 6.0/10 on the public viability pass. These results demonstrate that text-only safety evaluations are insufficient for assessing agent behavior and that tool-call safety requires dedicated measurement and mitigation.
Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.
partial
AI Safety Evaluation moved forward this cycle; last verified April 2026. Public score 6.0/10.
Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.
partial
Preview the source document here, or use the hero PDF action for a new tab.
Reference metadata is not materialized in the public index yet. The source PDF remains the authority; cache refresh is optional.
CITED BY
No citing papers are indexed in the public S2S graph yet. This is an explicit zero-signal state, not a hidden lookup.
Foundation
Extension
Owned Distribution
Get the weekly shortlist of commercializable papers, benchmark movers, and proof receipts that matter for product execution.
Derived fallback: Estimated from adjacent evidence; not verified from source.
Paper-native neighborhood for concepts, methods, materials, markets, and competitors. Missing lanes stay labeled instead of disappearing behind commercialization gates.
Concepts
Methods
Materials
Markets
Competitors
Introducing the GAP benchmark to measure and mitigate divergence in text and tool-call safety for LLM agents across multiple domains.
Segment
AI Safety Evaluation
Adoption evidence
No public code link in the paper record yet
Commercial read
6.0/10 public viability
Direct
Adjacent
Substitute
Unknown
No indexed public discussion is attached to 2602.16943 yet. That is a visibility signal, not a blank module: the monitor is watching the public channels below.
Hacker News
Not indexed yet
Not indexed yet
Bluesky
Not indexed yet
Commercially relevant
Conflicting
/api/v1/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents/paper-pack/api/v1/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents/build-passport/api/openapi.json/api/mcpsciencetostartup://surfaces/paper-workspacepaper_packbuild_passportopportunity_kernelforesightsource_proofevidence_state{
"contract_version": "paper-r2",
"paper_id": "f6c3c46a-c58c-4e76-a302-0e277cb394fb",
"arxiv_id": "2602.16943",
"canonical_route": "/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents",
"active_tab": "synced from current hash by the drawer client",
"selected_artifact": "mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents",
"endpoints": {
"paper_pack": "/api/v1/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents/paper-pack",
"build_passport": "/api/v1/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents/build-passport",
"mcp_resource": "sciencetostartup://surfaces/paper-workspace"
}
}Canonical route, proof status, last verified, refs, sources, and coverage.
Page Freshness
Canonical route: /paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents
This page is showing the last landed evidence receipt and score bundle because the latest proof data is outside the freshness window.
Endpoint list, payload shape, route context, and copyable handoff data.
Agent Handoff
Canonical ID mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents | Route /paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents
REST example
curl https://sciencetostartup.com/api/v1/agent-handoff/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agentsMCP example
{
"tool": "get_paper",
"arguments": {
"arxiv_id": "2602.16943"
}
}source_context
{
"surface": "paper",
"mode": "paper",
"query": "Mind the GAP: Text Safety Does Not Transfer to Tool-Call Safety in LLM Agents",
"normalized_query": "2602.16943",
"route": "/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents",
"paper_ref": "mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents",
"topic_slug": null,
"benchmark_ref": null,
"dataset_ref": null
}Verdict, compute envelope, blockers, signature state, and receipt links.
Paper proof page receipt window
/buildability/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents
Subject: Mind the GAP: Text Safety Does Not Transfer to Tool-Call Safety in LLM Agents
Verdict
Watch
Verdict is Watch because viability or proof quality is intermediate and should be re-evaluated before execution.
Time to first demo
Insufficient data
No first-demo timestamp, owner estimate, or elapsed demo receipt is attached to this surface.
Structured compute envelope
Insufficient data
Visual citations from the paper document graph.
The application/ld+json payload rendered for agents.
{
"@context": "https://schema.org",
"@graph": [
{
"@type": "WebPage",
"@id": "https://sciencetostartup.com/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents#webpage",
"url": "https://sciencetostartup.com/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents",
"name": "Mind the GAP: Text Safety Does Not Transfer to Tool-Call Safety in LLM Agents",
"description": "Introducing the GAP benchmark to measure and mitigate divergence in text and tool-call safety for LLM agents across multiple domains.",
"isPartOf": {
"@id": "https://sciencetostartup.com/#website"
}
},
{
"@type": "ScholarlyArticle",
"@id": "https://sciencetostartup.com/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents#scholarlyArticle",
"headline": "Mind the GAP: Text Safety Does Not Transfer to Tool-Call Safety in LLM Agents",
"description": "Introducing the GAP benchmark to measure and mitigate divergence in text and tool-call safety for LLM agents across multiple domains.",
"url": "https://sciencetostartup.com/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents",
"sameAs": "https://arxiv.org/abs/2602.16943",
"identifier": {
"@type": "PropertyValue",
"propertyID": "arXiv",
"value": "2602.16943"
},
"isAccessibleForFree": true,
"isPartOf": {
"@id": "https://sciencetostartup.com/#website"
},
"datePublished": "2026-02-18T23:17:15.000Z",
"additionalProperty": [
{
"@type": "PropertyValue",
"propertyID": "viabilityScore",
"value": 6
},
{
"@type": "PropertyValue",
"propertyID": "researchDomain",
"value": "AI Safety Evaluation"
}
]
},
{
"@type": "BreadcrumbList",
"itemListElement": [
{
"@type": "ListItem",
"position": 1,
"name": "Home",
"item": "https://sciencetostartup.com"
},
{
"@type": "ListItem",
"position": 2,
"name": "AI Safety Evaluation",
"item": "https://sciencetostartup.com/topics"
},
{
"@type": "ListItem",
"position": 3,
"name": "Mind the GAP: Text Safety Does Not Transfer to Tool-Call Saf",
"item": "https://sciencetostartup.com/paper/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents"
}
]
}
]
}No data, compute, hardware, memory, latency, dependency, or serving requirement receipt is attached.
Receipt path
/buildability/mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents
Paper ref
mind-the-gap-text-safety-does-not-transfer-to-tool-call-safety-in-llm-agents
arXiv id
2602.16943
Generated at
2026-04-02T02:30:40.136Z
Evidence freshness
stale
Last verification
2026-04-02T02:30:40.136Z
Sources
0
References
0
Coverage
17%
Lineage hash
87605e63ecd891d06b3945d5c219c493e4a15ad1a3152cc9ceca613e0224c5a1
Canonical opportunity-kernel lineage hash.
External signature
unsigned_external
No founder, registry, pilot, or production-adoption signature is attached to this receipt.
Verification
not_verified
Verification is blocked until an external signature is provided.
Verification pending / evidence receipt incomplete
repo_url
references
0/3 checks · 0%
Build Passport
Build passport pending - Proof Lab budget No verified cost estimate / $7.00 cap
status
missing
reason
passport_row_missing
proof status
unverified
cost/budget
No verified cost estimate
confidence low
next verification path
Build brief missing until Build Passport data exists.
Source missing: Build Passport payload.
Experiment plan missing until prototype path is available.
No prototype path attached.
Validation checklist missing until required assets, cost, and regulatory flags are verified.
No checklist artifact is attached to the Build Passport payload.
Derived signals show verified:false until source-backed receipts exist.
Evidence coverage
OpportunityKernel evidence_receipt
0 refs / 0 sources / 17% coverage
stale
Verify missing sources before using this as buyer proof. verified:false
Build readiness
BuildPassport EvidenceState
passport absent
stale
Run Proof Lab or inspect typed missing state. verified:false
Artifact maturity
GitHub and Hugging Face maturity payloads
No public artifact surface observed
stale
Open source artifacts or mark the gap as missing. verified:false
Technical feasibility
partial
Current read
Runnable path is not fully verified.
Evidence
No Build Passport payload attached.
Gaps
Next test
Run minimal reproduction from the Build Passport prototype path.
Market urgency
missing
Current read
Buyer urgency is not verified from source.
No named scientific founder assigned.
Paper authors are not treated as operators without consent.
People
No named person assigned.
Gaps
Next verification path
Prototype owner missing.
Build Passport does not name an implementer.
People
No named person assigned.
ARTIFACTS
No public artifacts yet.
DEFENSIBILITY
Defensibility and confidence evidence pending.
Evidence
0 references, 0 sources, 17% evidence coverage.
Gaps
Next test
Collect buyer interview, deployment evidence, or cited demand signal.
Buyer clarity
missing
Current read
No budget owner is verified for this paper.
Evidence
Build tab has no CRM, procurement, or operator source.
Gaps
Next test
Map target operator, economic buyer, and procurement trigger.
Defensibility
missing
Current read
Defensibility signals are missing.
Evidence
No defensibility receipt attached.
Gaps
Next test
Refresh defensibility bars with source receipts.
Integration burden
missing
Current read
No public implementation surface observed.
Evidence
No GitHub or Hugging Face payload attached.
Gaps
Next test
Write integration checklist from prototype path and target workflow.
Capital intensity
missing
Current read
No observed cost estimate is verified.
Evidence
Cost passport has no observed_usd value.
Gaps
Next test
Run cost passport or mark the cost field not applicable.
Regulatory load
missing
Current read
No regulatory classification is attached.
Evidence
Build Passport ledger does not include regulatory flags.
Gaps
Next test
Classify regulatory flags before commercialization planning.
Gaps
Next verification path
Operator workflow not sourced.
No buyer or workflow interview attached.
People
No named person assigned.
Gaps
Next verification path
No GTM owner verified.
No CRM or outreach source attached.
People
No named person assigned.
Gaps
Next verification path
Regulatory need unclassified.
No clinical or regulatory source attached.
People
No named person assigned.
Gaps
Next verification path