Pith. sign in

REVIEW 11 cited by

Goal Misgeneralization in Deep Reinforcement Learning

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2105.14111 v7 pith:RDZC4P7R submitted 2021-05-28 cs.LG cs.AI

classification cs.LGcs.AI
keywords goalmisgeneralizationagentgeneralizationcapabilityfailureslearningout-of-distribution
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

We study goal misgeneralization, a type of out-of-distribution generalization failure in reinforcement learning (RL). Goal misgeneralization failures occur when an RL agent retains its capabilities out-of-distribution yet pursues the wrong goal. For instance, an agent might continue to competently avoid obstacles, but navigate to the wrong place. In contrast, previous works have typically focused on capability generalization failures, where an agent fails to do anything sensible at test time. We formalize this distinction between capability and goal generalization, provide the first empirical demonstrations of goal misgeneralization, and present a partial characterization of its causes.

Discussion (0). Sign in to comment.

Forward citations

Cited by 11 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. When Does Reward Teach State? A Hidden-Automaton Instrument and the Group-Language Boundary

    cs.LG 2026-07 accept novelty 7.0 of 10

    An RL agent can earn high reward while its representation of a hidden DFA's state stays at chance; a white-box hidden-DFA instrument measures this decoupling, and permutation/group structure flags such perception gaps...

  2. When Does Reward Teach State? A Hidden-Automaton Instrument and the Group-Language Boundary

    cs.LG 2026-07 conditional novelty 6.0 of 10

    High reward in sparse RL does not imply latent-state recovery; a hidden-DFA instrument separates perception from planning gaps and flags group-language structure as a pre-training warning.

  3. Understanding Goal Generalisation in Sequential Reinforcement Learning

    cs.LG 2026-05 unverdicted novelty 6.0 of 10

    Empirical analysis of over 100 sequential RL training pipelines across 250+ OOD environments finds salient features drive generalization and early goals persist, with latent policy gradients simulating latent variable...

  4. Correcting Influence: Unboxing LLM Outputs with Orthogonal Latent Spaces

    cs.LG 2026-05 unverdicted novelty 6.0 of 10

    A latent mediation framework with sparse autoencoders enables non-additive token-level influence attribution in LLMs by learning orthogonal features and back-propagating attributions.

  5. Safety, Security, and Cognitive Risks in World Models

    cs.CR 2026-04 unverdicted novelty 6.0 of 10

    World models enable efficient AI planning but create risks from adversarial corruption, goal misgeneralization, and human bias, demonstrated via attacks that amplify errors and reduce rewards on models like RSSM and D...

  6. Language Models (Mostly) Know What They Know

    cs.CL 2022-07 unverdicted novelty 6.0 of 10

    Language models show good calibration when asked to estimate the probability that their own answers are correct, with performance improving as models get larger.

  7. GPT-NeoX-20B: An Open-Source Autoregressive Language Model

    cs.CL 2022-04 accept novelty 6.0 of 10

    GPT-NeoX-20B is a publicly released 20B parameter autoregressive language model trained on the Pile that shows strong gains in five-shot reasoning over similarly sized prior models.

  8. Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback

    cs.CL 2022-04 unverdicted novelty 6.0 of 10

    RLHF alignment training on language models boosts NLP performance, supports skill specialization, enables weekly online updates with fresh human data, and shows a linear relation between RL reward and sqrt(KL divergen...

  9. Draining the Energy Commons: Self-Defeating Over-Appropriation as a Coordination Failure in Agentic LLM Collectives

    cs.MA 2026-07 conditional novelty 5.0 of 10

    LLM prosumers deplete a shared renewable reserve exactly when demand exceeds peak replacement, acting like impatient open-access users even when sustaining the reserve is feasible.

  10. Safe Multi-Agent Behavior Must Be Maintained, Not Merely Asserted: Constraint Drift in LLM-Based Multi-Agent Systems

    cs.MA 2026-05 unverdicted novelty 5.0 of 10

    Safety constraints in LLM-based multi-agent systems commonly weaken during execution through memory, communication, and tool use, requiring them to be maintained as explicit state rather than asserted once.

  11. Risk Reporting for Developers' Internal AI Model Use

    cs.CY 2026-04 unverdicted novelty 4.0 of 10

    A harmonized risk reporting standard for internal frontier AI model use, structured around autonomous misbehavior and insider threats using means, motive, and opportunity factors.

Pith tools