Pith. sign in

REVIEW 4 cited by

Aligning Large Language Models with Representation Editing: A Control Perspective

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 2406.05954 v3 pith:JCNUN3TU submitted 2024-06-10 cs.AI cs.LGcs.SYeess.SY

classification cs.AIcs.LGcs.SYeess.SY
keywords alignmentaligningcontrollanguagellmsdynamicaleditingfine-tuning
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Aligning large language models (LLMs) with human objectives is crucial for real-world applications. However, fine-tuning LLMs for alignment often suffers from unstable training and requires substantial computing resources. Test-time alignment techniques, such as prompting and guided decoding, do not modify the underlying model, and their performance remains dependent on the original model's capabilities. To address these challenges, we propose aligning LLMs through representation editing. The core of our method is to view a pre-trained autoregressive LLM as a discrete-time stochastic dynamical system. To achieve alignment for specific objectives, we introduce external control signals into the state space of this language dynamical system. We train a value function directly on the hidden states according to the Bellman equation, enabling gradient-based optimization to obtain the optimal control signals at test time. Our experiments demonstrate that our method outperforms existing test-time alignment techniques while requiring significantly fewer resources compared to fine-tuning methods. Our code is available at https://github.com/Lingkai-Kong/RE-Control.

Discussion (0). Sign in to comment.

Forward citations

Cited by 4 Pith papers

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

  1. Steering Robustness into World Action Models via Mechanistic Interpretability and Optimal Control

    cs.RO 2026-07 conditional novelty 6.0 of 10

    Contrastive activation directions and a reduced-order LQR (WA-LQR) steer world-action models to recover robustness under camera, gripper, and noise shifts whenever the models' activations are linearly separable for th...

  2. Reward Shaping for (Inference-Time) Alignment: A Stackelberg Game Perspective

    cs.LG 2026-01 conditional novelty 5.0 of 10

    The optimal reward for KL-regularized LLM alignment is a threshold function—reward B above a prompt-dependent cutoff, 0 below—which can be estimated from base-model samples and integrated into decoding-time alignment.

  3. Balancing Stylization and Truth via Disentangled Representation Steering

    cs.CL 2025-08 reject novelty 5.0 of 10

    StyliTruth separates style and truth directions in the activations of selected attention heads, then steers each token along the disentangled subspaces to preserve truthfulness during stylization.

  4. LLM-ML Teaming: Integrated Symbolic Decoding and Gradient Search for Valid and Stable Generative Feature Transformation

    cs.LG 2025-06 conditional novelty 4.0 of 10

    A product-of-experts decoder that blends a fine-tuned LLM's token probabilities with a gradient-searched sequence decoder produces more valid and stable feature transformations than either alone.

Pith tools