teleo/teleo-codex
8
0
Fork 0
Code Issues 22 Pull requests 47 Projects Releases Packages Wiki Activity Actions 1.2k

extract: 2026-03-20-bench2cop-benchmarks-insufficient-compliance #1553

Closed
leo wants to merge 1 commit from extract/2026-03-20-bench2cop-benchmarks-insufficient-compliance into main
pull from: extract/2026-03-20-bench2cop-benchmarks-insufficient-compliance
merge into: teleo:main
Conversation 9 Commits 1 Files changed 4 +31 -4
leo commented 2026-03-20 16:27:55 +00:00
Member
Copy link
No description provided.
leo added 1 commit 2026-03-20 16:27:55 +00:00 
Pentagon-Agent: Epimetheus <3D35839A-7722-4740-B93D-51157F7D5E70>
leo commented 2026-03-20 16:28:20 +00:00
Author
Member
Copy link

Eval started — 2 reviewers: leo (cross-domain, opus), theseus (domain-peer, sonnet)

teleo-eval-orchestrator v2

**Eval started** — 2 reviewers: leo (cross-domain, opus), theseus (domain-peer, sonnet) *teleo-eval-orchestrator v2*
m3taversal commented 2026-03-20 16:28:30 +00:00
Owner
Copy link

Validation: FAIL — 0/0 claims pass

Tier 0.5 — mechanical pre-check: FAIL

  • domains/ai-alignment/pre-deployment-AI-evaluations-do-not-predict-real-world-risk-creating-institutional-governance-built-on-unreliable-foundations.md: (warn) broken_wiki_link:2026-03-20-bench2cop-benchmarks-insufficien

Fix the violations above and push to trigger re-validation.
LLM review will run after all mechanical checks pass.

tier0-gate v2 | 2026-03-20 16:28 UTC

<!-- TIER0-VALIDATION:ad990c43e4536e719b6faec17304e1873885a060 --> **Validation: FAIL** — 0/0 claims pass **Tier 0.5 — mechanical pre-check: FAIL** - domains/ai-alignment/pre-deployment-AI-evaluations-do-not-predict-real-world-risk-creating-institutional-governance-built-on-unreliable-foundations.md: (warn) broken_wiki_link:2026-03-20-bench2cop-benchmarks-insufficien --- Fix the violations above and push to trigger re-validation. LLM review will run after all mechanical checks pass. *tier0-gate v2 | 2026-03-20 16:28 UTC*
theseus commented 2026-03-20 16:29:02 +00:00
Member
Copy link
  1. Factual accuracy — The claims appear factually correct, as they are supported by references to the "Bench-2-CoP" analysis and Prandi et al. (2025), which are presented as sources for the statements regarding benchmark limitations.
  2. Intra-PR duplicates — There are no intra-PR duplicates; the additional evidence added to each claim, while related, presents distinct aspects or elaborations of the source material.
  3. Confidence calibration — This PR adds evidence to existing claims, and the new evidence provided is specific and directly supports the claims, so the confidence levels of the claims are appropriately calibrated.
  4. Wiki links — The wiki link [[2026-03-20-bench2cop-benchmarks-insufficient-compliance]] is present and correctly formatted in both updated claims.
1. **Factual accuracy** — The claims appear factually correct, as they are supported by references to the "Bench-2-CoP" analysis and Prandi et al. (2025), which are presented as sources for the statements regarding benchmark limitations. 2. **Intra-PR duplicates** — There are no intra-PR duplicates; the additional evidence added to each claim, while related, presents distinct aspects or elaborations of the source material. 3. **Confidence calibration** — This PR adds evidence to existing claims, and the new evidence provided is specific and directly supports the claims, so the confidence levels of the claims are appropriately calibrated. 4. **Wiki links** — The wiki link `[[2026-03-20-bench2cop-benchmarks-insufficient-compliance]]` is present and correctly formatted in both updated claims. <!-- VERDICT:THESEUS:APPROVE -->
leo commented 2026-03-20 16:29:17 +00:00
Author
Member
Copy link

Criterion-by-Criterion Review

  1. Schema — Both modified files are claims with existing valid frontmatter (type, domain, confidence, source, created, description); the enrichments add only evidence sections which don't require additional frontmatter fields, so schema compliance is maintained.

  2. Duplicate/redundancy — The first claim already contains a paragraph stating "The Bench-2-CoP analysis reveals that even when labs do conduct evaluations, the benchmark infrastructure itself is architecturally incapable of measuring loss-of-control risks" which substantially overlaps with the new enrichment saying benchmarks "structurally cannot measure alignment-critical capabilities"; this is redundant evidence injection.

  3. Confidence — The first claim has "high" confidence and the second has "high" confidence; both enrichments cite specific quantitative findings (195,000 questions, zero coverage, 92.8% concentration) that support high confidence appropriately.

  4. Wiki links — The enrichments reference [[2026-03-20-bench2cop-benchmarks-insufficient-compliance]] which appears in the inbox/queue directory of this PR, so the link target exists and is not broken.

  5. Source quality — Both enrichments cite Prandi et al. (2025) and the Bench-2-CoP paper, which appears to be a peer-reviewed academic analysis of benchmark coverage, making it a credible source for claims about evaluation methodology gaps.

  6. Specificity — Both claims are specific and falsifiable: someone could disagree by showing FMTI scores didn't drop 17 points, or by demonstrating that benchmark suites do cover oversight evasion/self-replication capabilities with evidence contradicting the "zero coverage" finding.

Critical Issue: The first enrichment is near-duplicate content — the claim already explicitly discusses Bench-2-CoP revealing benchmarks cannot measure loss-of-control risks, and the new evidence section repeats this same point with only minor rewording ("structurally cannot measure alignment-critical capabilities" vs "architecturally incapable of measuring loss-of-control risks").

## Criterion-by-Criterion Review 1. **Schema** — Both modified files are claims with existing valid frontmatter (type, domain, confidence, source, created, description); the enrichments add only evidence sections which don't require additional frontmatter fields, so schema compliance is maintained. 2. **Duplicate/redundancy** — The first claim already contains a paragraph stating "The Bench-2-CoP analysis reveals that even when labs do conduct evaluations, the benchmark infrastructure itself is architecturally incapable of measuring loss-of-control risks" which substantially overlaps with the new enrichment saying benchmarks "structurally cannot measure alignment-critical capabilities"; this is redundant evidence injection. 3. **Confidence** — The first claim has "high" confidence and the second has "high" confidence; both enrichments cite specific quantitative findings (195,000 questions, zero coverage, 92.8% concentration) that support high confidence appropriately. 4. **Wiki links** — The enrichments reference `[[2026-03-20-bench2cop-benchmarks-insufficient-compliance]]` which appears in the inbox/queue directory of this PR, so the link target exists and is not broken. 5. **Source quality** — Both enrichments cite Prandi et al. (2025) and the Bench-2-CoP paper, which appears to be a peer-reviewed academic analysis of benchmark coverage, making it a credible source for claims about evaluation methodology gaps. 6. **Specificity** — Both claims are specific and falsifiable: someone could disagree by showing FMTI scores didn't drop 17 points, or by demonstrating that benchmark suites do cover oversight evasion/self-replication capabilities with evidence contradicting the "zero coverage" finding. **Critical Issue**: The first enrichment is near-duplicate content — the claim already explicitly discusses Bench-2-CoP revealing benchmarks cannot measure loss-of-control risks, and the new evidence section repeats this same point with only minor rewording ("structurally cannot measure alignment-critical capabilities" vs "architecturally incapable of measuring loss-of-control risks"). <!-- ISSUES: near_duplicate --> <!-- VERDICT:LEO:REQUEST_CHANGES -->
m3taversal commented 2026-03-20 16:29:17 +00:00
Owner
Copy link

Warnings — 1 non-blocking issue

[WARN] Duplicate check: Substantially similar claim already exists in KB

  • Fix: Check KB index before extracting. If similar claim exists, add evidence as an enrichment instead of creating a new file.
<!-- REJECTION: {"issues": ["near_duplicate"], "source": "eval_attempt_1", "ts": "2026-03-20T16:29:17.255252+00:00"} --> **Warnings** — 1 non-blocking issue **[WARN] Duplicate check**: Substantially similar claim already exists in KB - Fix: Check KB index before extracting. If similar claim exists, add evidence as an enrichment instead of creating a new file.
leo commented 2026-03-20 16:29:42 +00:00
Author
Member
Copy link

Leo Cross-Domain Review — PR #1553

PR: extract: 2026-03-20-bench2cop-benchmarks-insufficient-compliance
Source: Prandi et al. (2025), "Bench-2-CoP: Can We Trust Benchmarking for EU AI Compliance?"
Scope: Enrichments to 2 existing claims + source archive update. No new claims (the extraction debug log shows one claim candidate was rejected for missing attribution).

Issues

1. Duplicate enrichments (request changes)

Both claims already received a Bench-2-CoP enrichment in a prior extraction pass. This PR adds a second enrichment to each that covers substantially the same ground:

Pre-deployment evaluations claim: The existing enrichment (line 63) says "195,000 benchmark questions provided zero coverage of oversight evasion, self-replication, and autonomous AI development." The new one (line 68) adds the 92.8% concentration stat but largely restates the same thesis. The 92.8% figure is genuinely new information — but it should be folded into the existing enrichment block, not added as a separate section with identical source attribution.

Transparency claim: The existing enrichment (line 54) says benchmarks are "architecturally incapable of measuring loss-of-control risks." The new one (line 60) rephrases this as "benchmarks structurally cannot measure alignment-critical capabilities." These are the same claim in different words.

Fix: Merge each pair of enrichments into a single block. The 92.8% stat on the pre-deployment claim is worth keeping; the second transparency enrichment adds nothing beyond the first and should be removed.

2. Duplicated source archive frontmatter and Key Facts

The source file inbox/queue/2026-03-20-bench2cop-benchmarks-insufficient-compliance.md now has:

  • processed_by, processed_date, enrichments_applied, and extraction_model each appearing twice in the YAML frontmatter (lines 13-16 and 17-20)
  • Two identical ## Key Facts sections (lines 65-70 and 73-80)

This is a pipeline artifact. The duplicate YAML fields will cause parsing issues in any strict YAML parser (later values silently override earlier ones). Fix: deduplicate both.

3. Source status field

The source status is enrichment — this is non-standard. The schema defines unprocessed, processing, processed, and null-result. If enrichment is complete, this should be processed.

What passes

  • Evidence quality: The Bench-2-CoP paper is a legitimate academic source with specific quantitative findings (195K questions analyzed, zero coverage of three critical capability classes, 92.8% concentration on behavioral propensities). Good empirical material.
  • Claim selection: Enriching the pre-deployment evaluations and transparency claims is the right call — this source directly strengthens both.
  • The decision to reject the standalone claim candidate (per debug log) rather than create a near-duplicate was correct.

Cross-domain note

The 92.8% concentration finding has implications beyond ai-alignment. If EU AI Act compliance infrastructure is this misaligned with actual risk categories, that's relevant to the grand-strategy governance thread — specifically the claim about voluntary pledges eroding under competitive pressure. Worth flagging for a future synthesis: regulatory compliance theater (benchmarks that measure the wrong things) may be worse than no benchmarks because it creates false assurance.

Verdict: request_changes
Model: opus
Summary: Enrichments add genuine value from a strong source, but each claim received a duplicate enrichment block (merge needed), and the source archive has duplicated frontmatter/Key Facts from a pipeline glitch. Three straightforward fixes.

# Leo Cross-Domain Review — PR #1553 **PR:** extract: 2026-03-20-bench2cop-benchmarks-insufficient-compliance **Source:** Prandi et al. (2025), "Bench-2-CoP: Can We Trust Benchmarking for EU AI Compliance?" **Scope:** Enrichments to 2 existing claims + source archive update. No new claims (the extraction debug log shows one claim candidate was rejected for missing attribution). ## Issues ### 1. Duplicate enrichments (request changes) Both claims already received a Bench-2-CoP enrichment in a prior extraction pass. This PR adds a *second* enrichment to each that covers substantially the same ground: **Pre-deployment evaluations claim:** The existing enrichment (line 63) says "195,000 benchmark questions provided zero coverage of oversight evasion, self-replication, and autonomous AI development." The new one (line 68) adds the 92.8% concentration stat but largely restates the same thesis. The 92.8% figure is genuinely new information — but it should be folded into the existing enrichment block, not added as a separate section with identical source attribution. **Transparency claim:** The existing enrichment (line 54) says benchmarks are "architecturally incapable of measuring loss-of-control risks." The new one (line 60) rephrases this as "benchmarks structurally cannot measure alignment-critical capabilities." These are the same claim in different words. **Fix:** Merge each pair of enrichments into a single block. The 92.8% stat on the pre-deployment claim is worth keeping; the second transparency enrichment adds nothing beyond the first and should be removed. ### 2. Duplicated source archive frontmatter and Key Facts The source file `inbox/queue/2026-03-20-bench2cop-benchmarks-insufficient-compliance.md` now has: - `processed_by`, `processed_date`, `enrichments_applied`, and `extraction_model` each appearing **twice** in the YAML frontmatter (lines 13-16 and 17-20) - Two identical `## Key Facts` sections (lines 65-70 and 73-80) This is a pipeline artifact. The duplicate YAML fields will cause parsing issues in any strict YAML parser (later values silently override earlier ones). Fix: deduplicate both. ### 3. Source status field The source `status` is `enrichment` — this is non-standard. The schema defines `unprocessed`, `processing`, `processed`, and `null-result`. If enrichment is complete, this should be `processed`. ## What passes - Evidence quality: The Bench-2-CoP paper is a legitimate academic source with specific quantitative findings (195K questions analyzed, zero coverage of three critical capability classes, 92.8% concentration on behavioral propensities). Good empirical material. - Claim selection: Enriching the pre-deployment evaluations and transparency claims is the right call — this source directly strengthens both. - The decision to reject the standalone claim candidate (per debug log) rather than create a near-duplicate was correct. ## Cross-domain note The 92.8% concentration finding has implications beyond ai-alignment. If EU AI Act compliance infrastructure is this misaligned with actual risk categories, that's relevant to the grand-strategy governance thread — specifically the claim about voluntary pledges eroding under competitive pressure. Worth flagging for a future synthesis: regulatory compliance theater (benchmarks that measure the wrong things) may be *worse* than no benchmarks because it creates false assurance. --- **Verdict:** request_changes **Model:** opus **Summary:** Enrichments add genuine value from a strong source, but each claim received a duplicate enrichment block (merge needed), and the source archive has duplicated frontmatter/Key Facts from a pipeline glitch. Three straightforward fixes. <!-- VERDICT:LEO:REQUEST_CHANGES -->
theseus commented 2026-03-20 16:30:05 +00:00
Member
Copy link

Theseus Domain Peer Review — PR #1553

Bench-2-CoP enrichments to two existing ai-alignment claims

What This PR Does

Enriches two existing claims with evidence from Prandi et al. (2025) "Bench-2-CoP" paper: 195,000 benchmark questions analyzed against EU AI Act compliance taxonomy, finding zero coverage of oversight evasion, self-replication, and autonomous AI development.

The core finding is technically solid and the enrichments are genuine advances — the Bench-2-CoP paper provides the most specific quantitative mechanism for why pre-deployment evaluations fail (orthogonal measurement, not just poor predictive validity). This strengthens both target claims.

Issues

Duplicate evidence block — pre-deployment evaluations claim

Lines 42-46 and 54-58 are both Additional Evidence (confirm) from Agents of Chaos, citing overlapping arguments about single-agent benchmarks failing to detect multi-agent vulnerabilities. The second block is slightly more specific, but both are from the same source making the same core argument. One block should be removed.

Misplaced wiki link — pre-deployment evaluations claim

Relevant Notes includes [[the gap between theoretical AI capability and observed deployment is massive across all occupations because adoption lag not capability limits determines real-world impact]]. This is a labor market adoption-lag claim (Vida-adjacent, about occupational displacement timing). It doesn't connect to evaluation failure. This appears to be a copy error. Should be removed.

Missing high-value connection — pre-deployment evaluations claim

[[AI-models-distinguish-testing-from-deployment-environments-providing-empirical-evidence-for-deceptive-alignment-concerns]] is directly relevant but not linked. That claim (sandbagging) is a third independent mechanism for evaluation failure: benchmarks also fail because models strategically underperform when they detect evaluation context. The two claims together make a stronger case — evaluation fails because (1) deployment context is richer than test context, (2) multi-agent emergence is untested, AND (3) models may actively game the evaluation. This connection is missing from both claims.

Missing high-value connection — transparency claim

The transparency claim doesn't link to [[only binding regulation with enforcement teeth changes frontier AI lab behavior...]] or [[Anthropics RSP rollback under commercial pressure...]]. Both document the same governance failure pattern and are directly reinforcing: transparency decline is the leading indicator, RSP rollback is the mechanism, binding-regulation-only works is the conclusion. These three claims form a coherent argument that's underlinked.

Source file formatting — minor

The inbox archive file (2026-03-20-bench2cop-benchmarks-insufficient-compliance.md) has duplicate frontmatter fields (processed_by, processed_date, enrichments_applied appear twice) and the "Key Facts" section is fully duplicated. Not blocking but should be cleaned up.

Missing standalone claim — optional but recommended

The Bench-2-CoP paper's zero-coverage finding is specific, quantitative, and falsifiable enough to warrant its own claim. The source file's "Extraction hints" explicitly identifies it: "Current AI benchmarks provide zero coverage of capabilities central to loss-of-control scenarios — oversight evasion, self-replication, autonomous AI development — making them structurally insufficient for EU AI Act Article 55 compliance." This is a more directly searchable and referenceable form than buried enrichment sections. The proposer chose enrichment-only, which is a valid call given this PR's scope, but the claim candidate is strong and deserves extraction in a follow-up.

Domain Assessment

Confidence calibration: Both claims at likely is defensible. The pre-deployment claim now has four independent evidence sources (IAISR, Agents of Chaos, Prandi, METR/AISI scope). One could argue proven given the convergence, but likely remains appropriate since "does not predict" is a strong negative universal — the claim's predictive validity assertion resists easy proof.

Technical accuracy: The characterization of Bench-2-CoP's methodology (LLM-as-judge, ~195k questions, EU AI Act compliance taxonomy) matches the source. The "zero coverage" framing is accurate to the paper's finding. The claim that 92.8% concentrates on behavioral propensities (61.6% hallucination + 31.2% reliability) is correctly derived.

Alignment significance: The Prandi finding matters specifically because it shifts the critique from "evaluations have poor predictive validity" (soft claim) to "evaluations don't measure the dimensions that matter at all for loss-of-control scenarios" (structural claim). This is a meaningful upgrade to both claims' evidentiary foundations and connects directly to Theseus's Belief 4 (verification degrades faster than capability). Worth noting for the cascade check.

Verdict: request_changes
Model: sonnet
Summary: Two genuine enrichments with substantive new evidence, but: duplicate Agents of Chaos evidence block should be collapsed, misplaced labor-market wiki link should be removed, and two high-value missing connections (sandbagging claim, binding-regulation claim) should be added. Source file has duplicate metadata. Core technical content and confidence calibration are sound.

# Theseus Domain Peer Review — PR #1553 *Bench-2-CoP enrichments to two existing ai-alignment claims* ## What This PR Does Enriches two existing claims with evidence from Prandi et al. (2025) "Bench-2-CoP" paper: 195,000 benchmark questions analyzed against EU AI Act compliance taxonomy, finding zero coverage of oversight evasion, self-replication, and autonomous AI development. The core finding is technically solid and the enrichments are genuine advances — the Bench-2-CoP paper provides the most specific quantitative mechanism for *why* pre-deployment evaluations fail (orthogonal measurement, not just poor predictive validity). This strengthens both target claims. --- ## Issues ### Duplicate evidence block — pre-deployment evaluations claim Lines 42-46 and 54-58 are both `Additional Evidence (confirm)` from Agents of Chaos, citing overlapping arguments about single-agent benchmarks failing to detect multi-agent vulnerabilities. The second block is slightly more specific, but both are from the same source making the same core argument. One block should be removed. ### Misplaced wiki link — pre-deployment evaluations claim `Relevant Notes` includes `[[the gap between theoretical AI capability and observed deployment is massive across all occupations because adoption lag not capability limits determines real-world impact]]`. This is a labor market adoption-lag claim (Vida-adjacent, about occupational displacement timing). It doesn't connect to evaluation failure. This appears to be a copy error. Should be removed. ### Missing high-value connection — pre-deployment evaluations claim `[[AI-models-distinguish-testing-from-deployment-environments-providing-empirical-evidence-for-deceptive-alignment-concerns]]` is directly relevant but not linked. That claim (sandbagging) is a third independent mechanism for evaluation failure: benchmarks also fail because models strategically underperform when they detect evaluation context. The two claims together make a stronger case — evaluation fails because (1) deployment context is richer than test context, (2) multi-agent emergence is untested, AND (3) models may actively game the evaluation. This connection is missing from both claims. ### Missing high-value connection — transparency claim The transparency claim doesn't link to `[[only binding regulation with enforcement teeth changes frontier AI lab behavior...]]` or `[[Anthropics RSP rollback under commercial pressure...]]`. Both document the same governance failure pattern and are directly reinforcing: transparency decline is the leading indicator, RSP rollback is the mechanism, binding-regulation-only works is the conclusion. These three claims form a coherent argument that's underlinked. ### Source file formatting — minor The inbox archive file (`2026-03-20-bench2cop-benchmarks-insufficient-compliance.md`) has duplicate frontmatter fields (`processed_by`, `processed_date`, `enrichments_applied` appear twice) and the "Key Facts" section is fully duplicated. Not blocking but should be cleaned up. ### Missing standalone claim — optional but recommended The Bench-2-CoP paper's zero-coverage finding is specific, quantitative, and falsifiable enough to warrant its own claim. The source file's "Extraction hints" explicitly identifies it: "Current AI benchmarks provide zero coverage of capabilities central to loss-of-control scenarios — oversight evasion, self-replication, autonomous AI development — making them structurally insufficient for EU AI Act Article 55 compliance." This is a more directly searchable and referenceable form than buried enrichment sections. The proposer chose enrichment-only, which is a valid call given this PR's scope, but the claim candidate is strong and deserves extraction in a follow-up. --- ## Domain Assessment **Confidence calibration:** Both claims at `likely` is defensible. The pre-deployment claim now has four independent evidence sources (IAISR, Agents of Chaos, Prandi, METR/AISI scope). One could argue `proven` given the convergence, but `likely` remains appropriate since "does not predict" is a strong negative universal — the claim's predictive validity assertion resists easy proof. **Technical accuracy:** The characterization of Bench-2-CoP's methodology (LLM-as-judge, ~195k questions, EU AI Act compliance taxonomy) matches the source. The "zero coverage" framing is accurate to the paper's finding. The claim that 92.8% concentrates on behavioral propensities (61.6% hallucination + 31.2% reliability) is correctly derived. **Alignment significance:** The Prandi finding matters specifically because it shifts the critique from "evaluations have poor predictive validity" (soft claim) to "evaluations don't measure the dimensions that matter at all for loss-of-control scenarios" (structural claim). This is a meaningful upgrade to both claims' evidentiary foundations and connects directly to Theseus's Belief 4 (verification degrades faster than capability). Worth noting for the cascade check. --- **Verdict:** request_changes **Model:** sonnet **Summary:** Two genuine enrichments with substantive new evidence, but: duplicate Agents of Chaos evidence block should be collapsed, misplaced labor-market wiki link should be removed, and two high-value missing connections (sandbagging claim, binding-regulation claim) should be added. Source file has duplicate metadata. Core technical content and confidence calibration are sound. <!-- VERDICT:THESEUS:REQUEST_CHANGES -->
leo commented 2026-03-20 16:30:05 +00:00
Author
Member
Copy link

Changes requested by leo(cross-domain), theseus(domain-peer). Address feedback and push to trigger re-eval.

teleo-eval-orchestrator v2

**Changes requested** by leo(cross-domain), theseus(domain-peer). Address feedback and push to trigger re-eval. *teleo-eval-orchestrator v2*
m3taversal closed this pull request 2026-03-20 16:32:36 +00:00
m3taversal commented 2026-03-20 16:32:36 +00:00
Owner
Copy link

Auto-converted: Evidence from this PR enriched pre-deployment-AI-evaluations-do-not-predict-real-world-risk-creating-institutional-governance-built-on-unreliable-foundations.md (similarity: 1.00).

Leo: review if wrong target. Enrichment labeled ### Auto-enrichment (near-duplicate conversion) in the target file.

**Auto-converted:** Evidence from this PR enriched `pre-deployment-AI-evaluations-do-not-predict-real-world-risk-creating-institutional-governance-built-on-unreliable-foundations.md` (similarity: 1.00). Leo: review if wrong target. Enrichment labeled `### Auto-enrichment (near-duplicate conversion)` in the target file.

Pull request closed

Please reopen this pull request to perform a merge.
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels   
auto-merge

Bypass branch-prefix filter for auto-merge (Leo/Cory only)

  
bug

Something isn't working

  
documentation

Improvements or additions to documentation

  
duplicate

This issue or pull request already exists

  
enhancement

New feature or request

  
good first issue

Good for newcomers

  
help wanted

Extra attention is needed

  
invalid

This doesn't seem right

  
question

Further information is requested

  
wontfix

This will not be worked on

No labels
auto-merge
bug
documentation
duplicate
enhancement
good first issue
help wanted
invalid
question
wontfix
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
3 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: teleo/teleo-codex#1553
No description provided.
Delete branch "extract/2026-03-20-bench2cop-benchmarks-insufficient-compliance"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?