Refusal in language models is mediated by a single direction in residual stream activations that can be erased to disable safety or added to elicit refusal.
Title resolution pending
8 Pith papers cite this work. Polarity classification is still indexing.
8
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
Benchmark construction artifacts in hallucination detection corpora allow naive text-similarity baselines to achieve near-perfect scores, and controlled evaluations show most methods perform near chance except SAPLMA and the new DRIFT probe.
Deep Pre-Alignment uses a small VLM perceiver instead of ViT to pre-align visual features with LLM text space, yielding 1.9-3.0 point gains on multimodal benchmarks and 32.9% less language forgetting.
BoostAPR boosts automated program repair by training a sequence-level assessor and line-level credit allocator from execution outcomes, then applying them in PPO to reach 40.7% on SWE-bench Verified.
The power distribution is the target of power sampling, the closed-form solution to self-reward KL-regularized RL, and the basis for power self-distillation that matches sampling performance at lower cost.
EAGLE resolves feature-level uncertainty in speculative sampling via one-step token advancement, delivering 2.7x-3.5x speedup on LLaMA2-Chat 70B and doubled throughput across multiple model families and tasks.
MAPLE uses meta-learning with prototypical networks to learn transferable representations and achieves state-of-the-art cross-prompt essay scoring on ELLIPSE, LAILA, and parts of ASAP datasets.
Representations learned by large AI models are converging toward a shared statistical model of reality.
citing papers explorer
-
Refusal in Language Models Is Mediated by a Single Direction
Refusal in language models is mediated by a single direction in residual stream activations that can be erased to disable safety or added to elicit refusal.
-
PARALLAX: Separating Genuine Hallucination Detection from Benchmark Construction Artifacts
Benchmark construction artifacts in hallucination detection corpora allow naive text-similarity baselines to achieve near-perfect scores, and controlled evaluations show most methods perform near chance except SAPLMA and the new DRIFT probe.
-
Deep Pre-Alignment for VLMs
Deep Pre-Alignment uses a small VLM perceiver instead of ViT to pre-align visual features with LLM text space, yielding 1.9-3.0 point gains on multimodal benchmarks and 32.9% less language forgetting.
-
BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models
BoostAPR boosts automated program repair by training a sequence-level assessor and line-level credit allocator from execution outcomes, then applying them in PPO to reach 40.7% on SWE-bench Verified.
-
Power Distribution Bridges Sampling, Self-Reward RL, and Self-Distillation
The power distribution is the target of power sampling, the closed-form solution to self-reward KL-regularized RL, and the basis for power self-distillation that matches sampling performance at lower cost.
-
EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty
EAGLE resolves feature-level uncertainty in speculative sampling via one-step token advancement, delivering 2.7x-3.5x speedup on LLaMA2-Chat 70B and doubled throughput across multiple model families and tasks.
-
MAPLE: A Meta-learning Framework for Cross-Prompt Essay Scoring
MAPLE uses meta-learning with prototypical networks to learn transferable representations and achieves state-of-the-art cross-prompt essay scoring on ELLIPSE, LAILA, and parts of ASAP datasets.
-
The Platonic Representation Hypothesis
Representations learned by large AI models are converging toward a shared statistical model of reality.