Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning | Paper Citation Record

Observation ab2705ea-ad33-47eb-825f-c2e3b31a29df · outbound

This paper cites and Zhang, Edwin , year =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning and Zhang, Edwin , year =

Reference 1

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:25.817060Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2abb9b30-a985-4b00-943f-e9301274ae68 · outbound

This paper cites , year =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning , year =

Reference 2

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:25.888007Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 57826263-223c-4742-be08-037d3faa7ccf · outbound

This paper cites Beyond Binary Rewards: Training LMs to Reason About Their Uncertainty.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Beyond Binary Rewards: Training LMs to Reason About Their Uncertainty

Reference 3

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:26.000469Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 676c5148-e63d-4158-bbf2-c2f8b97722f6 · outbound

This paper cites TruthRL: Incentivizing Truthful LLMs via Reinforcement Learning.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning TruthRL: Incentivizing Truthful LLMs via Reinforcement Learning

Reference 4

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:26.113437Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1d0ec2be-5923-4e84-8d32-c4160d977077 · outbound

This paper cites TIAR: Trajectory-Informed Advantage Reweighting for LLM Abstention Learning.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning TIAR: Trajectory-Informed Advantage Reweighting for LLM Abstention Learning

Reference 5

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:26.206601Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 54305e50-0fb7-46ae-be6e-c870cb08f6e0 · outbound

This paper cites KARL: Mitigating Hallucinations in LLMs via Knowledge-Boundary-Aware Reinforcement Learning.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning KARL: Mitigating Hallucinations in LLMs via Knowledge-Boundary-Aware Reinforcement Learning

Reference 6

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:26.307965Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 04ea5d40-19af-43f6-9220-e1a19046f1ea · outbound

This paper cites UCPO: Uncertainty-Aware Policy Optimization.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning UCPO: Uncertainty-Aware Policy Optimization

Reference 7

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:26.423440Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation fef8b9ba-3f0d-43a2-adfa-7c7b669dd7bb · outbound

This paper cites Abstain-R1: Calibrated Abstention and Post-Refusal Clarification via Verifiable RL.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Abstain-R1: Calibrated Abstention and Post-Refusal Clarification via Verifiable RL

Reference 8

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:26.556359Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3f7d78d5-7b3a-4d41-9312-00c744b3dfb1 · outbound

This paper cites Enhancing Reliability across Short and Long-Form.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Enhancing Reliability across Short and Long-Form

Reference 9

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:26.707407Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 74247269-072c-4a8b-a849-8dbaae796dad · outbound

This paper cites Rejection Improves Reliability: Training LLMs to Refuse Unknown Questions Using RL from Knowledge Feedback.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Rejection Improves Reliability: Training LLMs to Refuse Unknown Questions Using RL from Knowledge Feedback

Reference 10

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:26.819594Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation fdec0010-df81-43b6-9070-93d785f61194 · outbound

This paper cites The Hallucination Tax of Reinforcement Finetuning , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning The Hallucination Tax of Reinforcement Finetuning , eprint =

Reference 11

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:26.884202Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d5138077-ffb5-40e6-9fb9-6dc1749a82c3 · outbound

This paper cites 2509.17730 , archivePrefix =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning 2509.17730 , archivePrefix =

Reference 12

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:26.972189Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ef39490a-612b-4557-a6b6-ad575cf526ed · outbound

This paper cites Knowledge-Level Consistency Reinforcement Learning: Dual-Fact Alignment for Long-Form Factuality , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Knowledge-Level Consistency Reinforcement Learning: Dual-Fact Alignment for Long-Form Factuality , eprint =

Reference 13

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:27.061604Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c3f7dae9-8e02-4c1d-b9b8-aee43e2b6828 · outbound

This paper cites 2505.13529 , archivePrefix =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning 2505.13529 , archivePrefix =

Reference 14

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:27.195630Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9ca9606d-b173-4b24-a852-b638efab5d88 · outbound

This paper cites Vanishing Gradients in Reinforcement Finetuning of Language Models , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Vanishing Gradients in Reinforcement Finetuning of Language Models , eprint =

Reference 15

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:27.338083Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 48d7edbf-492b-4f17-ad60-91b93f9125f3 · outbound

This paper cites On the Global Convergence Rates of Softmax Policy Gradient Methods.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning On the Global Convergence Rates of Softmax Policy Gradient Methods

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:27.440871Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c4689bc2-5a92-442e-bde5-03b62c4c55e5 · outbound

This paper cites The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models , eprint =

Reference 17

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:27.529142Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 05b2a897-3b9e-419b-aadc-8f88a138c232 · outbound

This paper cites Understanding the Effects of RLHF on LLM Generalisation and Diversity.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Understanding the Effects of RLHF on LLM Generalisation and Diversity

Reference 18

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:27.653093Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2b1af7e4-cad4-4e02-8915-1aa5cb3720c6 · outbound

This paper cites A Theory of Regularized Markov Decision Processes.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning A Theory of Regularized Markov Decision Processes

Reference 19

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:27.851497Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 352f964e-166d-4f4f-8e61-4f50a47f5763 · outbound

This paper cites Leverage the Average: an Analysis of KL Regularization in RL.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Leverage the Average: an Analysis of KL Regularization in RL

Reference 20

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:27.989253Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5cb1418d-f3ba-4e48-abb8-6c1abf01582c · outbound

This paper cites DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models

Reference 21

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:28.131447Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 34d3b5f7-0649-4dcc-af3c-139723f97f59 · outbound

This paper cites Understanding R1-Zero-Like Training: A Critical Perspective.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Understanding R1-Zero-Like Training: A Critical Perspective

Reference 22

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:28.214601Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ab733218-cc76-46b6-9b54-775408158617 · outbound

This paper cites Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMs.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMs

Reference 23

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:28.299798Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b4773189-261c-489d-8eb2-2ad401d27a4e · outbound

This paper cites DAPO: An Open-Source LLM Reinforcement Learning System at Scale.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning DAPO: An Open-Source LLM Reinforcement Learning System at Scale

Reference 24

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:28.428593Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2f5698f8-f765-4fbe-a238-c6f77a46ffbd · outbound

This paper cites REINFORCE++: Stabilizing Critic-Free Policy Optimization with Global Advantage Normalization.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning REINFORCE++: Stabilizing Critic-Free Policy Optimization with Global Advantage Normalization

Reference 25

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:28.530409Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4a2fb1da-8723-4123-9238-55217ffb70fc · outbound

This paper cites Scaling Laws for Reward Model Overoptimization , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Scaling Laws for Reward Model Overoptimization , eprint =

Reference 26

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:28.668683Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation f2005d49-85be-4b61-a798-a3125ee9fea5 · outbound

This paper cites an unresolved cited work.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Unresolved cited work

Reference 27

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:28.847058Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5cea912c-8fb8-426c-9511-ada9ed84a4ae · outbound

This paper cites and Martic, Miljan and others , year =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning and Martic, Miljan and others , year =

Reference 28

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:28.926687Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation da372744-d219-44ed-9d85-ee50eaf7f8d6 · outbound

This paper cites and others , year =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning and others , year =

Reference 29

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:28.981204Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6a66bdae-0c77-466d-93b0-ff8b2785630e · outbound

This paper cites Training Language Models to Follow Instructions with Human Feedback , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Training Language Models to Follow Instructions with Human Feedback , eprint =

Reference 30

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:29.084946Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 94218d1a-a73b-44d1-9a4f-e0c056e33543 · outbound

This paper cites an unresolved cited work.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Unresolved cited work

Reference 31

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:29.148434Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d9ebbbce-2630-4dc0-83ea-267f841fa17e · outbound

This paper cites On the Foundations of Noise-free Selective Classification , journal =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning On the Foundations of Noise-free Selective Classification , journal =

Reference 32

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:29.203476Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 312536c2-1da6-43f1-b6f6-0a97df50dd84 · outbound

This paper cites Selective Classification for Deep Neural Networks , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Selective Classification for Deep Neural Networks , eprint =

Reference 33

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:29.208690Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b6d3ec95-38a8-4e46-92a4-1d6507a674c2 · outbound

This paper cites Selective Question Answering under Domain Shift , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Selective Question Answering under Domain Shift , eprint =

Reference 34

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:29.275043Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 84154b10-a95b-4b4a-b71b-8a7515f75373 · outbound

This paper cites Out-of-Distribution Detection and Selective Generation for Conditional Language Models , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Out-of-Distribution Detection and Selective Generation for Conditional Language Models , eprint =

Reference 35

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:29.487503Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 82234b30-2e58-45fa-8300-7f711880624d · outbound

This paper cites Language Models (Mostly) Know What They Know , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Language Models (Mostly) Know What They Know , eprint =

Reference 36

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:29.603289Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5fb6dfa4-3519-424a-b12a-94c35fdf8497 · outbound

This paper cites Teaching Models to Express Their Uncertainty in Words , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Teaching Models to Express Their Uncertainty in Words , eprint =

Reference 37

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:29.696163Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3793a836-01b9-40e1-b150-26d2d3d9b44a · outbound

This paper cites Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedback , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedback , eprint =

Reference 38

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:29.836202Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6bb6a822-fbde-45a1-920e-d56aa29c8c99 · outbound

This paper cites Reported Confidence in LLMs Tracks Commitment More Than Correctness.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Reported Confidence in LLMs Tracks Commitment More Than Correctness

Reference 39

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:29.955373Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2f7a097b-94e0-402c-8c92-2de6209a4dc2 · outbound

This paper cites , year =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning , year =

Reference 40

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.086006Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ae44adc7-85fd-42af-afc3-f4f6f889dd0e · outbound

This paper cites Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation , eprint =

Reference 41

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.187212Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation e3e7f2be-5139-4f2a-b3f4-b3c0afd1ee4c · outbound

This paper cites Mitigating LLM Hallucinations via Conformal Abstention.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Mitigating LLM Hallucinations via Conformal Abstention

Reference 42

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.253798Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation dc84d3f3-b641-4ecf-89fd-0dea84786c69 · outbound

This paper cites Qwen2.5 Technical Report.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Qwen2.5 Technical Report

Reference 43

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.337364Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 894f0cf3-5f82-4d9f-8219-47967dfb7db4 · outbound

This paper cites TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension

Reference 44

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.434651Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9a01facf-b339-482a-9fbf-418f418d2f16 · outbound

This paper cites When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memories , eprint =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memories , eprint =

Reference 45

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.529942Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 24ce766d-cdd2-4133-991c-c12bf47fb26d · outbound

This paper cites 2511.13029 , archivePrefix =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning 2511.13029 , archivePrefix =

Reference 46

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.589609Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c5a2ff5f-f9dc-425f-acd1-3dc3e29d3f8c · outbound

This paper cites , year =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning , year =

Reference 47

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.694160Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 264e1248-4b57-4457-b282-25a819bc00a1 · outbound

This paper cites , year =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning , year =

Reference 48

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.769359Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 8f3000b0-2145-4245-b08d-d9a984197b8a · outbound

This paper cites , year =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning , year =

Reference 49

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.816895Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d3efdc51-f862-4439-8739-f1cf7d1215df · outbound

This paper cites , year =.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning , year =

Reference 50

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.876999Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 681405dc-2711-4b67-a25b-d529ac63e90d · outbound

This paper cites Approximating.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning Approximating

Reference 51

Resolution: unresolved; no resolver link, observed 2026-08-04T00:53:30.969336Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Coverage vector

Reference resolution

Outbound references

Pith citing papers