Lying Is Just a Phase: The Hidden Alignment Transition in Language Model Scaling

• URL: http://arxiv.org/abs/2605.18838v1
• Date: Wed, 13 May 2026 03:14:09 GMT
• Title: Lying Is Just a Phase: The Hidden Alignment Transition in Language Model Scaling
• Abstract summary: We measure the coupling between reasoning and truthfulness across 63 base models from 16 families.<n>We find a regime change invisible to loss curves below a family-dependent critical scale $N_c$, capabilities anticorrelate; above it, they cooperate.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Scaling laws predict loss from compute but not how capabilities interact. We measure the coupling between reasoning and truthfulness across 63 base models from 16 families and find a regime change invisible to loss curves: below a family-dependent critical scale $N_c$, capabilities anticorrelate; above it, they cooperate. $N_c \approx 3.5$B parameters [2.9B, 13.4B] (bootstrap 95% CI), but model size is not the only variable that determines phase. Architecture, data curation, and training recipe each shift $N_c$ independently: curated training eliminated the coupling dip between Qwen generations ($0.025 \to 0.830$ at matched scale), Gemma-4 at 4B achieves coupling 0.871, characteristic of 13B+ standard-trained models, through distillation and architectural innovation, and Phi at 1B matches web-trained coupling at 10B through data curation alone. Width normalization eliminates the anticorrelation across all tested families, supporting an output-projection bottleneck. Internally, 38 of 40 models show zero competing attention heads. A sparse-regression ODE cross-predicts held-out Llama-2 at 5.6% error. The diagnostic requires no model internals -- only public benchmark scores across a model family. The cooperative regime extends to the frontier ($r = +0.72$, 34 models, 10 labs). Code, data, and an open-source activation-steering tool for any open-weight model are released alongside an interactive dashboard that diagnoses any model's coupling phase, suggests concrete interventions (data curation, width, benchmark rotation), and provides ODE scaling predictions, frontier diagnostics, and eigenstructure analysis: https://zehenlabs.com/cape/.

Related papers

• Scaling Laws for Classical Machine Learning on Tabular Data: A Benchmark Study [0.0]
  127 graduate students each ran a fixed protocol on 3 assigned datasets.<n>For 5 of 6 families, a single family-level exponent predicts each family's cross-dataset curves nearly as well as per-dataset exponents.<n>We release the aggregated curves, per-cell fits, and a practical data-requirement table for N* to reach target error 0.15.
  arXiv  Detail & Related papers  (2026-07-23T23:45:09Z)
• Scaffolding the Strategist: Architecture-Dependent Reasoning Interventions in Hotelling Spatial Markets [0.01269104766024433]
  We investigate whether structured reasoning interventions improve the strategic economic reasoning of large language models.<n>We evaluate GPT-4.1-mini (a standard instruction-following model) and GPT-5-mini (a reasoning-optimized model) under five conditions.<n>We find a statistically significant crossover interaction between scaffolding type and model architecture.
  arXiv  Detail & Related papers  (2026-07-03T10:23:31Z)
• Unextractable Protocol Models: Collaborative Training and Inference without Weight Materialization [58.14514930760722]
  We consider a decentralized setup in which the participants collaboratively train and serve a large neural network.<n>In this setup, we explore the possibility of unmaterializable weights, where a full weight set is never available to any one participant.<n>We introduce Unextractable Protocol Models (UPMs): a training and inference framework that leverages the sharded model setup.
  arXiv  Detail & Related papers  (2026-05-22T10:24:57Z)
• Weight Decay Regimes in Grokking Transformers: Cheap Online Diagnostics [0.0]
  Transformers trained on modular arithmetic exhibit sharp transitions between memorization, generalization, and collapse.<n>We show that weight decay acts as a scalar empirical control parameter for these regimes.<n>We introduce online diagnostics, mean pairwise attentionhead cosine similarity and entropy standard deviation, that track training dynamics.
  arXiv  Detail & Related papers  (2026-05-19T19:48:40Z)
• The Growing Pains of Frontier Models: When Leaderboards Stop Separating and What to Measure Next [0.0]
  Leaderboards rank frontier models on independent axes but do not reveal whether capabilities reinforce or trade off across releases.<n>We decompose paired SWE-bench and GPQA Diamond scores into a population coupling trend and per-release residual.<n>We provide a three-level playbook (locate, diagnose, rotate), a per-lab measurement-priority table, and seven falsifiable predictions.
  arXiv  Detail & Related papers  (2026-05-13T03:19:38Z)
• The Geometry of Forgetting: Temporal Knowledge Drift as an Independent Axis in LLM Representations [50.43168858368539]
  Large language models confidently produce outdated answers, and no existing method can detect them.<n>We show this is not an engineering failure but a structural one: temporal drift is encoded as a direction in the residual stream geometrically to both correctness and uncertainty.
  arXiv  Detail & Related papers  (2026-05-09T22:27:31Z)
• Compact SO(3) Equivariant Atomistic Foundation Models via Structural Pruning [4.793059213046564]
  We propose a structural pruning method for SO(3) equivariant atomistic foundation models.<n>The pruning is applied along the channel and order dimensions, with each irreducible representation kept or removed as a complete block.<n>The pruned MACE-MP model outperforms the official from-scratch trained small model on 7 of 9 metrics on the Matbench Discovery leaderboard.
  arXiv  Detail & Related papers  (2026-05-09T11:07:03Z)
• Three Roles, One Model: Role Orchestration at Inference Time to Close the Performance Gap Between Small and Large Agents [0.4666493857924357]
  We study whether inference-time scaffolding alone, without any additional training compute, can improve the performance of a small model in complex multi-step environments.<n>We evaluate Qwen3-8B on the AppWorld benchmark under both full-precision and 4-bit quantized configurations.<n>On full-precision inference, our scaffolded 8B model surpasses DeepSeek-Coder 33B Instruct (7.1%) from the original AppWorld evaluation.
  arXiv  Detail & Related papers  (2026-04-13T13:40:33Z)
• Learn by Surprise, Commit by Proof [0.0]
  We propose a self-gated post-training framework for autonomous knowledge acquisition.<n>When a passage produces anomalously high pertoken loss, L SCP flags it, generates a Q&A chain that forces the model to articulate its own knowledge.<n>The entire learning intensity is governed by a single parameter $r$.
  arXiv  Detail & Related papers  (2026-04-02T12:17:10Z)
• MERGETUNE: Continued fine-tuning of vision-language models [77.8627788911249]
  Fine-tuning vision-language models (VLMs) often leads to catastrophic forgetting of pretrained knowledge.<n>We introduce a novel paradigm, continued fine-tuning (CFT), which seeks to recover pretrained knowledge after a zero-shot model has already been adapted.
  arXiv  Detail & Related papers  (2026-01-15T15:15:53Z)
• Silhouette-based Gait Foundation Model [56.27974816297294]
  Building a unified gait foundation model requires addressing two longstanding barriers: Scalability and Generalization.<n>We introduce FoundationGait, the first scalable, self-supervised pretraining framework for gait understanding.
  arXiv  Detail & Related papers  (2025-11-30T01:53:41Z)
• Language models scale reliably with over-training and on downstream tasks [121.69867718185125]
  Scaling laws are useful guides for derisking expensive training runs. However, there remain gaps between current studies and how language models are trained. In contrast, scaling laws mostly predict loss on inference, but models are usually compared on downstream task performance.
  arXiv  Detail & Related papers  (2024-03-13T13:54:00Z)
• Tangent Model Composition for Ensembling and Continual Fine-tuning [69.92177580782929]
  Tangent Model Composition (TMC) is a method to combine component models independently fine-tuned around a pre-trained point. TMC improves accuracy by 4.2% compared to ensembling non-linearly fine-tuned models.
  arXiv  Detail & Related papers  (2023-07-16T17:45:33Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.