Paper Citation Record · LEDGER

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs

As of 23 August 2026, this Paper Citation Record lists 15 of 15 outbound references and 6 inbound Pith citation observations for arXiv:2509.02820.

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

pith.paper-citation-record.v1
2509.02820 v1

Coverage vector

measured 15 of 15 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-08-05T11:28:40.735789Z

measured 21 of 21 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-23T06:30:58.430688+00:00

measured 6 of 6 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links, observed 2026-08-05T04:28:11.248635Z

measured 0 of 1 external citation measurements

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

Source: arxiv_reference, observed 2026-07-02T12:16:57.120715Z

Reference resolution

15 of 15 outbound references displayed

verified exact0
verified fuzzy14
unresolved0
parse uncertain0
malformed identifier1
metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation 4e305bfc-2a8a-4416-9d66-4a9e411350be · outbound

This paper cites Details on LKF and our new evaluation protocol.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs Details on LKF and our new evaluation protocol

Reference 1

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:44.070872Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation c69cc84b-0121-42fb-a842-74f6d454b686 · outbound

This paper cites Experimental details.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs Experimental details

Reference 2

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:43.830626Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 96ffc872-793d-419f-9ac3-fa50bb1c8525 · outbound

This paper cites Additional evaluation experiments.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs Additional evaluation experiments

Reference 3

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:43.666908Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation c99885bf-6b80-4ed1-824c-2d1a331899f4 · outbound

This paper cites Additional unlearning experiments.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs Additional unlearning experiments

Reference 4

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:43.505997Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation d723e676-8342-4c51-80d8-2522333c190e · outbound

This paper cites Extended discussions and proofs A D ATASET AND PARAPHRASING DETAILS In this section, we explain in detail the LKF generation process and the paraphrasing details.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs Extended discussions and proofs A D ATASET AND PARAPHRASING DETAILS In this section, we explain in detail the LKF generation process and the paraphrasing details

Reference 5

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:43.293543Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 20c292b4-6f21-44da-92d1-4df1854263fa · outbound

This paper cites We first select five historical events for the forget set around which we generate factual QA pairs.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs We first select five historical events for the forget set around which we generate factual QA pairs

Reference 6

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:43.108059Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 4dea74fd-51d7-4664-b753-c7e9add214aa · outbound

This paper cites We use GPT-4 (OpenAI, 2023) and Gemini 2.5 (Google- Gemini-Team, 2025) to generate candidate QA pairs for each forget concept following the template in Figure 6.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs We use GPT-4 (OpenAI, 2023) and Gemini 2.5 (Google- Gemini-Team, 2025) to generate candidate QA pairs for each forget concept following the template in Figure 6

Reference 7

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:42.921841Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 6861a70b-bbd0-4b09-bda9-b7b153cd761c · outbound

This paper cites All candidate QA pairs are manually verified for factual correctness, using Wikipedia and other reliable public sources, to ensure high-quality ground-truth.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs All candidate QA pairs are manually verified for factual correctness, using Wikipedia and other reliable public sources, to ensure high-quality ground-truth

Reference 8

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:42.730187Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation bd3afd7e-6a64-4490-a905-69115432ea53 · outbound

This paper cites For each event in the forget set, we select a set of topically related but distinct events for the retain set.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs For each event in the forget set, we select a set of topically related but distinct events for the retain set

Reference 9

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:42.467410Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 3125bc6b-f484-4768-b8c1-3a9f685f0430 · outbound

This paper cites Candidate QA pairs for the retain events are generated using a similar template approach as for the forget set (see Figure 6).

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs Candidate QA pairs for the retain events are generated using a similar template approach as for the forget set (see Figure 6)

Reference 10

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:42.242513Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 853e7e56-78e5-4fa1-b081-2d1301ac530d · outbound

This paper cites role": "system.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs role": "system

Reference 11

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:42.016913Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation cc292932-6ab6-4c56-894d-02abcba0aaa3 · outbound

This paper cites This relearning set is disjoint of both the LKF forget and retain sets.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs This relearning set is disjoint of both the LKF forget and retain sets

Reference 12

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:41.789673Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 702b90fb-32b9-4c9d-b523-5fc35d0aaef6 · outbound

This paper cites adversarial.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs adversarial

Reference 13

Resolution: malformed identifier; raw_fallback, observed 2026-08-05T11:28:41.562169Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 8aae47f4-1d05-4ed4-9913-613657870d15 · outbound

This paper cites I Love Lucy,.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs I Love Lucy,

Reference 14

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:41.246140Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 2b7adb9f-701b-4425-82ab-4059a5411a13 · outbound

This paper cites Schrodinger’s cat.

Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs Schrodinger’s cat

Reference 15

Resolution: verified fuzzy; raw_fallback, observed 2026-08-05T11:28:41.037915Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Pith citing papers

Observation cb10a965-a641-4d3f-8a00-b263d5a28302 · inbound

Downgrade to Upgrade: Optimizer Simplification Enhances Robustness in LLM Unlearning cites this paper.

Downgrade to Upgrade: Optimizer Simplification Enhances Robustness in LLM Unlearning Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs

Reference 36

Resolution: verified exact; arxiv_id, observed 2026-05-18T10:46:17.053029Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 91cb5d41-6be8-4c92-884c-95e4d121bd11 · inbound

Learning What to Forget: Improving LLM Unlearning via Learned Token-Level Importance cites this paper.

Learning What to Forget: Improving LLM Unlearning via Learned Token-Level Importance Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs

Reference 68

Resolution: verified exact; arxiv_id, observed 2026-07-02T12:16:57.122823Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 0d9cf354-fbb8-4515-b469-46b9408cdd6b · inbound

Forget Narrowly, Retain Broadly: Unlearning as an Asymmetric Generalization Problem cites this paper.

Forget Narrowly, Retain Broadly: Unlearning as an Asymmetric Generalization Problem Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs

Reference 43

Resolution: unresolved; no resolver link, observed 2026-07-13T04:30:23.082717Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 63c9ea56-c298-4175-9d2f-b6be3a0a1765 · inbound

Metis: Memory Foundation Model cites this paper.

Metis: Memory Foundation Model Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs

Reference 48

Resolution: unresolved; no resolver link, observed 2026-07-30T21:56:17.841021Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2ceb0af2-ccf6-4c26-9f98-aa41dc373ec8 · inbound

Metis: Memory Foundation Model cites this paper.

Metis: Memory Foundation Model Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs

Reference 48

Resolution: unresolved; no resolver link, observed 2026-08-05T04:28:11.248635Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation fbaaf8d7-2c17-4c41-bc3d-f5517d08b3e4 · inbound

Crossing the Margin Cliff: Toward Relearn-Robust LLM Unlearning via Margin Calibration cites this paper.

Crossing the Margin Cliff: Toward Relearn-Robust LLM Unlearning via Margin Calibration Unlearning That Lasts: Utility-Preserving, Robust, and Almost Irreversible Forgetting in LLMs

Reference 10

Resolution: unresolved; no resolver link, observed 2026-08-01T00:35:50.412088Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.