Copertina

The Boy Who Survived: Removing Harry Potter from an LLM is harder than reported

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autore principale:	Shostack, Adam
Natura:	Preprint
Pubblicazione:	2024
Soggetti:	Computation and Language Artificial Intelligence Machine Learning
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Harry Potter is Still Here! Probing Knowledge Leakage in Targeted Unlearned Large Language Models via Automated Adversarial Prompting
di: To, Bang Trinh Tran, et al.
Pubblicazione: (2025)

Who can we trust? LLM-as-a-jury for Comparative Assessment
di: Qian, Mengjie, et al.
Pubblicazione: (2026)

The illusion of a perfect metric: Why evaluating AI's words is harder than it looks
di: Oliva, Maria Paz, et al.
Pubblicazione: (2025)

Who Gets Which Message? Auditing Demographic Bias in LLM-Generated Targeted Text
di: Islam, Tunazzina
Pubblicazione: (2026)

Robustly Improving LLM Fairness in Realistic Settings via Interpretability
di: Karvonen, Adam, et al.
Pubblicazione: (2025)

Does Machine Unlearning Truly Remove Knowledge?
di: Chen, Haokun, et al.
Pubblicazione: (2025)

Mechanism of Task-oriented Information Removal in In-context Learning
di: Cho, Hakaze, et al.
Pubblicazione: (2025)

Language models are better than humans at next-token prediction
di: Shlegeris, Buck, et al.
Pubblicazione: (2022)

R.I.P.: Better Models by Survival of the Fittest Prompts
di: Yu, Ping, et al.
Pubblicazione: (2025)

AdapterSwap: Continuous Training of LLMs with Data Removal and Access-Control Guarantees
di: Fleshman, William, et al.
Pubblicazione: (2024)

Repeat After Me: Transformers are Better than State Space Models at Copying
di: Jelassi, Samy, et al.
Pubblicazione: (2024)

BiasGym: A Simple and Generalizable Framework for Analyzing and Removing Biases through Elicitation
di: Islam, Sekh Mainul, et al.
Pubblicazione: (2025)

BED-LLM: Intelligent Information Gathering with LLMs and Bayesian Experimental Design
di: Choudhury, Deepro, et al.
Pubblicazione: (2025)

Two are better than one: Context window extension with multi-grained self-injection
di: Han, Wei, et al.
Pubblicazione: (2024)

Revisiting Chain-of-Thought Prompting: Zero-shot Can Be Stronger than Few-shot
di: Cheng, Xiang, et al.
Pubblicazione: (2025)

Mechanistic origins of catastrophic forgetting: why RL preserves circuits better than SFT?
di: Nunez, Jeanmely Rojas, et al.
Pubblicazione: (2026)

Beyond the Last Answer: Your Reasoning Trace Uncovers More than You Think
di: Hammoud, Hasan Abed Al Kader, et al.
Pubblicazione: (2025)

Iterative Label Refinement Matters More than Preference Optimization under Weak Supervision
di: Ye, Yaowen, et al.
Pubblicazione: (2025)

Time-To-Inconsistency: A Survival Analysis of Large Language Model Robustness to Adversarial Attacks
di: Li, Yubo, et al.
Pubblicazione: (2025)

Mitigating LLM Hallucinations via Conformal Abstention
di: Yadkori, Yasin Abbasi, et al.
Pubblicazione: (2024)

MoEs Are Stronger than You Think: Hyper-Parallel Inference Scaling with RoE
di: Zibakhsh, Soheil, et al.
Pubblicazione: (2025)

Bridging the Gap: Enhancing LLM Performance for Low-Resource African Languages with New Benchmarks, Fine-Tuning, and Cultural Adjustments
di: Alhanai, Tuka, et al.
Pubblicazione: (2024)

OptiMer: Optimal Distribution Vector Merging Is Better than Data Mixing for Continual Pre-Training
di: Song, Haiyue, et al.
Pubblicazione: (2026)

Who Benefits From Sinus Surgery? Comparing Generative AI and Supervised Machine Learning for Predicting Surgical Outcomes in Chronic Rhinosinusitis
di: Chowdhury, Sayeed Shafayet, et al.
Pubblicazione: (2026)

LLM Cyber Evaluations Don't Capture Real-World Risk
di: Lukošiūtė, Kamilė, et al.
Pubblicazione: (2025)

IITK at SemEval-2024 Task 10: Who is the speaker? Improving Emotion Recognition and Flip Reasoning in Conversations via Speaker Embeddings
di: Patel, Shubham, et al.
Pubblicazione: (2024)

Set-LLM: A Permutation-Invariant LLM
di: Egressy, Beni, et al.
Pubblicazione: (2025)

LLM Chemistry Estimation for Multi-LLM Recommendation
di: Sanchez, Huascar, et al.
Pubblicazione: (2025)

Removing Spurious Correlation from Neural Network Interpretations
di: Fotouhi, Milad, et al.
Pubblicazione: (2024)

User-LLM: Efficient LLM Contextualization with User Embeddings
di: Ning, Lin, et al.
Pubblicazione: (2024)

Reactive Transformer (RxT) -- Stateful Real-Time Processing for Event-Driven Reactive Language Models
di: Filipek, Adam
Pubblicazione: (2025)

LogProber: Disentangling confidence from contamination in LLM responses
di: Yax, Nicolas, et al.
Pubblicazione: (2024)

VBART: The Turkish LLM
di: Turker, Meliksah, et al.
Pubblicazione: (2024)

LazyLLM: Dynamic Token Pruning for Efficient Long Context LLM Inference
di: Fu, Qichen, et al.
Pubblicazione: (2024)

BPO: Staying Close to the Behavior LLM Creates Better Online LLM Alignment
di: Xu, Wenda, et al.
Pubblicazione: (2024)

Large Language Models Assume People are More Rational than We Really are
di: Liu, Ryan, et al.
Pubblicazione: (2024)

LLM Maybe LongLM: Self-Extend LLM Context Window Without Tuning
di: Jin, Hongye, et al.
Pubblicazione: (2024)

LLM See, LLM Do: Guiding Data Generation to Target Non-Differentiable Objectives
di: Shimabucoro, Luísa, et al.
Pubblicazione: (2024)

MixEval: Deriving Wisdom of the Crowd from LLM Benchmark Mixtures
di: Ni, Jinjie, et al.
Pubblicazione: (2024)

R-Zero: Self-Evolving Reasoning LLM from Zero Data
di: Huang, Chengsong, et al.
Pubblicazione: (2025)