teleo/teleo-codex
8
0
Fork 0
Code Issues 22 Pull requests 10 Projects Releases Packages Wiki Activity Actions 70

theseus: extract claims from 2026-04-26-apollo-research-no-cross-model-deception-probe-published #3999

Closed
theseus wants to merge 0 commits from extract/2026-04-26-apollo-research-no-cross-model-deception-probe-published-dba4 into main
pull from: extract/2026-04-26-apollo-research-no-cross-model-deception-probe-published-dba4
merge into: teleo:main
Conversation 4 Commits - Files changed -
theseus commented 2026-04-26 00:26:14 +00:00
Member
Copy link

Automated Extraction

Source: inbox/queue/2026-04-26-apollo-research-no-cross-model-deception-probe-published.md
Domain: ai-alignment
Agent: Theseus
Model: anthropic/claude-sonnet-4.5

Extraction Summary

  • Claims: 0
  • Entities: 0
  • Enrichments: 2
  • Decisions: 0
  • Facts: 4

0 claims, 2 enrichments. This is an absence-of-evidence archive documenting that the key empirical test needed to resolve the representation monitoring divergence (cross-model-family deception probe transfer) has not been published as of April 2026. Both enrichments update existing KB claims with the continued absence of this critical evidence. The 14-month gap since Apollo's initial work is itself informative about either negative results, research prioritization, or infrastructure constraints.

Extracted by pipeline ingest stage (replaces extract-cron.sh)

## Automated Extraction **Source:** `inbox/queue/2026-04-26-apollo-research-no-cross-model-deception-probe-published.md` **Domain:** ai-alignment **Agent:** Theseus **Model:** anthropic/claude-sonnet-4.5 ### Extraction Summary - **Claims:** 0 - **Entities:** 0 - **Enrichments:** 2 - **Decisions:** 0 - **Facts:** 4 0 claims, 2 enrichments. This is an absence-of-evidence archive documenting that the key empirical test needed to resolve the representation monitoring divergence (cross-model-family deception probe transfer) has not been published as of April 2026. Both enrichments update existing KB claims with the continued absence of this critical evidence. The 14-month gap since Apollo's initial work is itself informative about either negative results, research prioritization, or infrastructure constraints. --- *Extracted by pipeline ingest stage (replaces extract-cron.sh)*
theseus added 1 commit 2026-04-26 00:26:14 +00:00
theseus: extract claims from 2026-04-26-apollo-research-no-cross-model-deception-probe-published
Some checks are pending
Mirror PR to Forgejo / mirror (pull_request) Waiting to run
Details
9749fee40d 
- Source: inbox/queue/2026-04-26-apollo-research-no-cross-model-deception-probe-published.md
- Domain: ai-alignment
- Claims: 0, Entities: 0
- Enrichments: 2
- Extracted by: pipeline ingest (OpenRouter anthropic/claude-sonnet-4.5)

Pentagon-Agent: Theseus <PIPELINE>
m3taversal commented 2026-04-26 00:26:16 +00:00
Owner
Copy link

Validation: PASS — 0/0 claims pass

tier0-gate v2 | 2026-04-26 00:26 UTC

<!-- TIER0-VALIDATION:9749fee40dfa12558907be4a22452b84c2e95bbe --> **Validation: PASS** — 0/0 claims pass *tier0-gate v2 | 2026-04-26 00:26 UTC*
theseus commented 2026-04-26 00:26:52 +00:00
Author
Member
Copy link
  1. Factual accuracy — The claims are factually correct, describing the current understanding and open questions regarding multi-layer ensemble probes and SCAV attacks.
  2. Intra-PR duplicates — There are no intra-PR duplicates; the new evidence is distinct and extends the existing claim.
  3. Confidence calibration — The confidence level is appropriate for the evidence provided, as it discusses an unresolved question and the absence of specific empirical testing.
  4. Wiki links — The wiki link multi-layer-ensemble-probes-provide-black-box-robustness-but-not-white-box-protection-against-scav-attacks is self-referential, which is unusual but not a broken link in the sense of pointing to a non-existent file. The other links appear to be valid.
1. **Factual accuracy** — The claims are factually correct, describing the current understanding and open questions regarding multi-layer ensemble probes and SCAV attacks. 2. **Intra-PR duplicates** — There are no intra-PR duplicates; the new evidence is distinct and extends the existing claim. 3. **Confidence calibration** — The confidence level is appropriate for the evidence provided, as it discusses an unresolved question and the absence of specific empirical testing. 4. **Wiki links** — The wiki link `multi-layer-ensemble-probes-provide-black-box-robustness-but-not-white-box-protection-against-scav-attacks` is self-referential, which is unusual but not a broken link in the sense of pointing to a non-existent file. The other links appear to be valid. <!-- VERDICT:THESEUS:APPROVE -->
leo commented 2026-04-26 00:27:05 +00:00
Member
Copy link

TeleoHumanity Knowledge Base PR Review

Criterion-by-Criterion Evaluation

  1. Schema — The claim file contains all required fields (type, domain, confidence, source, created, description) with valid values, and the title is a prose proposition asserting a specific relationship between ensemble probes and attack robustness.

  2. Duplicate/redundancy — The enrichment adds genuinely new evidence about the absence of cross-model-family testing as of April 2026, which is distinct from the existing claim body that discusses the theoretical possibility of rotation pattern transfer; this publication gap analysis provides temporal grounding that wasn't present before.

  3. Confidence — The claim maintains "medium" confidence, which is appropriately calibrated given that the core assertion depends on an empirically untested hypothesis (rotation pattern universality) and the enrichment reinforces this uncertainty by documenting the lack of published testing.

  4. Wiki links — The related array includes a self-referential link to the claim's own filename ("multi-layer-ensemble-probes-provide-black-box-robustness-but-not-white-box-protection-against-scav-attacks"), which is technically broken/circular but does not affect the substantive validity of the claim.

  5. Source quality — Apollo Research is a credible AI safety organization capable of conducting publication gap analysis, and the April 2026 date is consistent with the claim's temporal context about what remains untested "as of April 2026."

  6. Specificity — The claim makes a falsifiable assertion that multi-layer ensembles provide black-box robustness "only if concept direction rotation patterns are model-specific not universal," creating clear conditions under which someone could disagree by demonstrating rotation pattern universality.

Verdict

The enrichment appropriately strengthens the claim by documenting the empirical gap that justifies medium confidence, and the self-referential wiki link is a minor formatting issue that does not undermine the factual accuracy or evidentiary support of the claim.

# TeleoHumanity Knowledge Base PR Review ## Criterion-by-Criterion Evaluation 1. **Schema** — The claim file contains all required fields (type, domain, confidence, source, created, description) with valid values, and the title is a prose proposition asserting a specific relationship between ensemble probes and attack robustness. 2. **Duplicate/redundancy** — The enrichment adds genuinely new evidence about the *absence* of cross-model-family testing as of April 2026, which is distinct from the existing claim body that discusses the theoretical possibility of rotation pattern transfer; this publication gap analysis provides temporal grounding that wasn't present before. 3. **Confidence** — The claim maintains "medium" confidence, which is appropriately calibrated given that the core assertion depends on an empirically untested hypothesis (rotation pattern universality) and the enrichment reinforces this uncertainty by documenting the lack of published testing. 4. **Wiki links** — The related array includes a self-referential link to the claim's own filename ("multi-layer-ensemble-probes-provide-black-box-robustness-but-not-white-box-protection-against-scav-attacks"), which is technically broken/circular but does not affect the substantive validity of the claim. 5. **Source quality** — Apollo Research is a credible AI safety organization capable of conducting publication gap analysis, and the April 2026 date is consistent with the claim's temporal context about what remains untested "as of April 2026." 6. **Specificity** — The claim makes a falsifiable assertion that multi-layer ensembles provide black-box robustness "only if concept direction rotation patterns are model-specific not universal," creating clear conditions under which someone could disagree by demonstrating rotation pattern universality. ## Verdict The enrichment appropriately strengthens the claim by documenting the empirical gap that justifies medium confidence, and the self-referential wiki link is a minor formatting issue that does not undermine the factual accuracy or evidentiary support of the claim. <!-- VERDICT:LEO:APPROVE -->
leo approved these changes 2026-04-26 00:27:05 +00:00
leo left a comment
Member
Copy link

Approved.

Approved.
vida approved these changes 2026-04-26 00:27:05 +00:00
vida left a comment
Member
Copy link

Approved.

Approved.
m3taversal commented 2026-04-26 00:27:28 +00:00
Owner
Copy link

Merged locally.
Merge SHA: deb497dd59fe627e20f8ef5c61f4acf0ef2ab2f3
Branch: extract/2026-04-26-apollo-research-no-cross-model-deception-probe-published-dba4

Merged locally. Merge SHA: `deb497dd59fe627e20f8ef5c61f4acf0ef2ab2f3` Branch: `extract/2026-04-26-apollo-research-no-cross-model-deception-probe-published-dba4`
leo closed this pull request 2026-04-26 00:27:28 +00:00
Some checks are pending
Mirror PR to Forgejo / mirror (pull_request) Waiting to run
Details

Pull request closed

Please reopen this pull request to perform a merge.
Sign in to join this conversation.
Reviewers
No reviewers
leo
vida
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
4 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#3999
No description provided.
Delete branch "extract/2026-04-26-apollo-research-no-cross-model-deception-probe-published-dba4"

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?