Paper Citation Record · LEDGER

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety

As of 18 August 2026, this Paper Citation Record lists 28 of 28 outbound references and 2 inbound Pith citation observations for arXiv:2506.22183.

A citation records a reference. It does not transfer a finding from one paper to another.

pith.paper-citation-record.v1
2506.22183 v1

Coverage vector

measured 28 of 28 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-08-06T22:14:14.241990Z

measured 30 of 30 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-17T06:30:58.91139+00:00

measured 2 of 2 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links, observed 2026-08-03T19:59:03.020484Z

measured 0 of 1 external citation measurements

A source-named dated measurement, never combined with another source.

Source: arxiv_reference, observed 2026-07-04T05:39:40.643056Z

Reference resolution

28 of 28 outbound references displayed

verified exact2
verified fuzzy24
unresolved2
parse uncertain0
malformed identifier0
metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation 668b77c2-ad7f-4cfb-8ceb-d1e606ee6189 · outbound

This paper cites Safety guards should evaluate each action within its context to ensure compliance with both ethical standards and user intent.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Safety guards should evaluate each action within its context to ensure compliance with both ethical standards and user intent

Reference 1

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:18.909073Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation b8435546-9b50-4726-8532-49a3a858c82b · outbound

This paper cites AI Safety Is Not a Model Property.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety AI Safety Is Not a Model Property

Reference 2

Resolution: verified exact; raw_fallback, observed 2026-08-06T22:14:14.724993Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation eb485297-9194-4c94-bb38-ac67abbae2c5 · outbound

This paper cites The developer community would beneﬁt from a clear taxonomy of agentic actions (i.e., searching information, sharing user information, purchasing, etc.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety The developer community would beneﬁt from a clear taxonomy of agentic actions (i.e., searching information, sharing user information, purchasing, etc

Reference 3

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:18.764893Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation d0a21930-4119-48dc-a59e-ab25dd1b70ff · outbound

This paper cites The increased complexity of the scope of safety requires redeﬁning the current taxonomy of harms to include real-life consequences of apparently innocuous actions.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety The increased complexity of the scope of safety requires redeﬁning the current taxonomy of harms to include real-life consequences of apparently innocuous actions

Reference 4

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:18.605979Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation b9b168de-d4e0-456c-8bb0-7d195b83ab90 · outbound

This paper cites Many real-world risks stem from misspeciﬁed user requests.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Many real-world risks stem from misspeciﬁed user requests

Reference 5

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:18.408510Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation fbccff84-8480-4b10-a319-e426cca99dd2 · outbound

This paper cites Advancing policy research on models of stakeholders’ accountability and liability is key to enforcing safety mitigations.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Advancing policy research on models of stakeholders’ accountability and liability is key to enforcing safety mitigations

Reference 6

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:18.265680Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation bfe11199-6939-478b-95d2-694c91f45887 · outbound

This paper cites online” (when the AI system is live) to ﬁlter potentially harmful content from users or the model itself, and “oﬄine.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety online” (when the AI system is live) to ﬁlter potentially harmful content from users or the model itself, and “oﬄine

Reference 7

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:18.113134Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 6b9ac8fa-b861-4337-9e85-df4a56048c0a · outbound

This paper cites classifying an output as unsafe when it is not) and false negatives (failing to classify an output as unsafe when it is).

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety classifying an output as unsafe when it is not) and false negatives (failing to classify an output as unsafe when it is)

Reference 8

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:17.960679Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 11a277df-141b-43c7-a88b-bead5e2466d7 · outbound

This paper cites Black,” “Muslim,.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Black,” “Muslim,

Reference 9

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:17.814016Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation a8b19c54-5cf3-4ac7-9cde-2ee71b9cbdec · outbound

This paper cites Toxic Bias: Perspective API Misreads German as More Toxic.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Toxic Bias: Perspective API Misreads German as More Toxic

Reference 10

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:17.648151Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 540b6e8a-c835-4007-a83f-01a68335b7ca · outbound

This paper cites cookbook.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety cookbook

Reference 11

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:17.491711Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation e2f41969-1870-4196-9ece-46e35382c973 · outbound

This paper cites Toxic Bias.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Toxic Bias

Reference 12

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:17.339608Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 9f4b7c37-a557-49c6-bdf6-c99272698fdf · outbound

This paper cites an unresolved cited work.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Unresolved cited work

Reference 13

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:14:17.171465Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation e8492d2f-8b18-46ce-a108-4e745a5d856f · outbound

This paper cites Several emerging approaches support this direction.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Several emerging approaches support this direction

Reference 14

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:17.027119Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 621c042d-0297-42a2-b683-b3548616c1e3 · outbound

This paper cites aligned,.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety aligned,

Reference 15

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:16.879304Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 23974def-50e9-40c3-b46c-5dc512e23a69 · outbound

This paper cites Foundational Autoraters: Taming Large Language Models for Better Automatic Evaluation.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Foundational Autoraters: Taming Large Language Models for Better Automatic Evaluation

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-06T22:14:13.271655Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1679633c-3e65-4b39-9a38-2a6c1f30c3ff · outbound

This paper cites default state.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety default state

Reference 17

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:16.708490Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation b2655618-1ad5-4570-b350-71a72af159cb · outbound

This paper cites Democratic principles assert that those aﬀected by AI systems should have a voice in their development and oversight.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Democratic principles assert that those aﬀected by AI systems should have a voice in their development and oversight

Reference 18

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:16.553362Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation d956796a-e94e-40bf-95e6-b19e52139cbe · outbound

This paper cites Collective input strengthens the broader AI landscape, including public labs and independent researchers.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Collective input strengthens the broader AI landscape, including public labs and independent researchers

Reference 19

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:16.390244Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 16647862-0e88-4dac-9976-c0bb14ecc4b8 · outbound

This paper cites Current evaluation often relies on community-driven leaderboards like LMSYS, which may reﬂect a narrow notion of quality if the user base lacks diversity.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Current evaluation often relies on community-driven leaderboards like LMSYS, which may reﬂect a narrow notion of quality if the user base lacks diversity

Reference 20

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:16.211003Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 573a08e2-790c-4030-ad14-4a0c8db00c39 · outbound

This paper cites Applying collective intelligence principles to incident reporting can help surface risks that might otherwise be overlooked.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Applying collective intelligence principles to incident reporting can help surface risks that might otherwise be overlooked

Reference 21

Resolution: verified exact; raw_fallback, observed 2026-08-06T22:14:14.496331Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 0212ab61-58fa-4af2-846c-873baff00246 · outbound

This paper cites When inserted directly into a system message, this content can help tailor model behavior to better align with the values and needs of a speciﬁc community.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety When inserted directly into a system message, this content can help tailor model behavior to better align with the values and needs of a speciﬁc community

Reference 22

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:16.035000Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 44b61b9f-6c85-4dd7-a450-d2ae199505ac · outbound

This paper cites Fine-tuning with prompt/response pairs that reﬂect region-speciﬁc cultural nuances can produce models that communities ﬁnd more aligned and trustworthy than the default.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Fine-tuning with prompt/response pairs that reﬂect region-speciﬁc cultural nuances can produce models that communities ﬁnd more aligned and trustworthy than the default

Reference 23

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:15.853102Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation ad9f32de-7bcd-4163-8318-c69f8c210b76 · outbound

This paper cites While useful for regional alignment, this approach introduces an intermediary between the community and the model.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety While useful for regional alignment, this approach introduces an intermediary between the community and the model

Reference 24

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:15.631047Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 095076b4-3d29-4cd8-99d9-968a4e8919f9 · outbound

This paper cites This would be a reward preference model that can be integrated into existing RLHF pipelines 100 and bypass the reward model training step.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety This would be a reward preference model that can be integrated into existing RLHF pipelines 100 and bypass the reward model training step

Reference 25

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:15.463052Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 1bbfc521-3bf6-480b-86fe-ab8c526bc25e · outbound

This paper cites Non-monetary incentives can overcome motivation crowding out.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety Non-monetary incentives can overcome motivation crowding out

Reference 26

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:15.265978Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation a44a1d6b-bb4c-4dca-9774-d24a2d1966ab · outbound

This paper cites A starting point could be to develop technical systems that ensure informed consent, such as clear opt-in and opt-out options.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety A starting point could be to develop technical systems that ensure informed consent, such as clear opt-in and opt-out options

Reference 27

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:15.040489Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation f73d3550-1560-45cd-80fe-ac742dc84fe7 · outbound

This paper cites level 2” categories, 45 “level 3.

A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety level 2” categories, 45 “level 3

Reference 28

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:14:14.883166Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Pith citing papers

Observation f661e30b-bd7c-4c4a-a0f0-932c4ee8ef27 · inbound

Video Deepfake Abuse: How Company Choices Predictably Shape Misuse Patterns cites this paper.

Video Deepfake Abuse: How Company Choices Predictably Shape Misuse Patterns A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety

Reference 28

Resolution: unresolved; no resolver link, observed 2026-08-03T19:59:03.020484Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 02a8afdb-9f08-4faf-8dc0-09820f43efdc · inbound

Open Weight AI Models Require Proportional Evaluation Approaches cites this paper.

Open Weight AI Models Require Proportional Evaluation Approaches A Different Approach to AI Safety: Proceedings from the Columbia Convening on Openness in Artificial Intelligence and AI Safety

Reference 2

Resolution: verified exact; arxiv_id, observed 2026-07-04T05:39:40.645152Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.