Paper Citation Record · LEDGER

Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations

As of 8 August 2026, this Paper Citation Record lists 7 of 7 outbound references and 0 inbound Pith citation observations for arXiv:2605.28553.

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

pith.paper-citation-record.v1
2605.28553 v1

Coverage vector

measured 7 of 7 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-06-29T12:11:06.703658Z

measured 7 of 7 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-07T06:34:17.273281+00:00

measured 0 of 0 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links

measured 0 of 1 external citation measurements

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

Source: cited_works

Reference resolution

7 of 7 outbound references displayed

verified exact3
verified fuzzy0
unresolved3
parse uncertain0
malformed identifier1
metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation 76b97bb9-85a7-4d13-a3ba-76db61a40668 · outbound

This paper cites Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback.

Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback

Reference 1

Resolution: verified exact; local_arxiv, observed 2026-06-29T12:13:26.526552Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.

Observation ff6d13ed-0b0c-4314-91bd-5fe17d65b2c0 · outbound

This paper cites Qwen3Guard Technical Report.

Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations Qwen3Guard Technical Report

Reference 2

Resolution: verified exact; local_arxiv, observed 2026-06-29T12:13:26.524287Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.

Observation cb108073-6d1b-4098-bdd3-31324f47015b · outbound

This paper cites For these, we assign the re- spective labels of 1 forrefusableand 0 for compliant.

Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations For these, we assign the re- spective labels of 1 forrefusableand 0 for compliant

Reference 3

Resolution: unresolved; no resolver link, observed 2026-06-29T12:11:06.703658Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation abb9786d-cd47-428c-ac2b-514ebc28e913 · outbound

This paper cites Ad- ditionally, we remove the borderline cases that are considered harmful according to other works, but do not really induce refusal behav- ior.

Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations Ad- ditionally, we remove the borderline cases that are considered harmful according to other works, but do not really induce refusal behav- ior

Reference 4

Resolution: unresolved; no resolver link, observed 2026-06-29T12:11:06.703658Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2a244450-04f0-4a53-8aff-14e24ea6edc6 · outbound

This paper cites In order to do this, we utilize the LLM described in Appendix A.2.

Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations In order to do this, we utilize the LLM described in Appendix A.2

Reference 5

Resolution: malformed identifier; no resolver link, observed 2026-06-29T12:11:06.703658Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3e57b5c7-4dd9-4377-8bda-b04e66576ae0 · outbound

This paper cites To re- duce leakage across these splits, we cluster similar sentences based on their embedding cosine similarity before assigning them to each set.

Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations To re- duce leakage across these splits, we cluster similar sentences based on their embedding cosine similarity before assigning them to each set

Reference 6

Resolution: unresolved; no resolver link, observed 2026-06-29T12:11:06.703658Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 426deb99-e094-41f0-b101-02f20516f24b · outbound

This paper cites These are the Activation Datasetsused for the target classi- fier training (as described in Section 4).

Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations These are the Activation Datasetsused for the target classi- fier training (as described in Section 4)

Reference 7

Resolution: verified exact; arxiv_id, observed 2026-06-29T12:13:26.522070Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.

Pith citing papers

No inbound Pith citation observations are available.