arXiv Machine Learning
Jul 30
Constitutional Midtraining: Content Presence Drives Alignment Gains
arXiv:2607. 26654v1 Announce Type: cross Abstract: Post-training alignment is often shallow, eroding under fine-tuning.
Stress-testing Alignment Midtraining
The paper "Stress-testing Alignment Midtraining" examines the effectiveness of alignment midtraining (AMT), a technique that continues pretraining on alignment-relevant data to improve generalisation beyond post‑training methods. Experiments on models up to 110 billion parameters and 1 billion midtraining tokens reveal that AMT can steer a model’s motivation in simple scenarios, but its effects are quickly overridden by even a tiny fraction of finetuning data with a competing motivation. The study also shows that rule-following requires demonstrations in either the midtraining or post‑training datasets to be robustly learned, leading the authors to conclude that current public evidence is insufficient to confirm that AMT resolves the core alignment challenges of powerful AI systems.
Related stories
arXiv Machine Learning
Jul 30
Shared SFT Lessons Across Alignment, Model Organisms, and Toy Models
arXiv:2607. 26173v1 Announce Type: new Abstract: Alignment training, model organisms, and toy models are usually treated as separate research areas.
arXiv AI
2d ago
Alignment Forecasting: Predicting Misalignment From Training Data
The paper introduces Alignment Forecasting, a method for predicting whether fine‑tuning a language model on a given dataset will increase specific alignment failures such as deception or sycophancy. It presents ALIGNMENTFORECASTBENCH, a benchmark of over 5,000 forecasting questions across many models, datasets, and failure modes, and shows that a simple forecasting scaffold using an LLM’s assessment of dataset bias can outperform baseline forecasters. The authors demonstrate that filtering out high‑risk training examples identified by the forecaster can improve alignment in multiple‑choice evaluations, though benefits in open‑ended conversations remain uncertain.
arXiv Machine Learning
Jul 16
Value Drifts: Tracing Value Alignment During LLM Post-Training
arXiv:2510. 26707v2 Announce Type: replace-cross Abstract: As LLMs occupy an increasingly important role in society, they are more and more confronted with questions that require them not only to draw on their general knowledge but also to align with certain human value systems.
arXiv AI
Aug 13
AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses
arXiv:2608. 12307v1 Announce Type: cross Abstract: Recent work on distillation transfers the capabilities of large models to smaller ones often by updating the latter's parameters, through teacher forcing, on-policy distillation, and related training-time methods.
arXiv AI
Jun 10
What Fits (Into Few Tokens) Doesn't Overfit: Compression and Generalization in ML Research Agents
arXiv:2606. 11045v1 Announce Type: new Abstract: Reusing a held-out benchmark adaptively should, in principle, invite overfitting.
Hugging Face Trending Papers
Jul 28
Shared SFT Lessons Across Alignment, Model Organisms, and Toy Models
Alignment training, model organisms, and toy models are usually treated as separate research areas. But projects in all three frequently use supervised fine-tuning (SFT) to pursue the same underlying goals.
arXiv AI
2d ago
Trajectory Soup: Pushing the Compute-Scaling Frontier of LLM Mid-training via Diverse Trajectories
arXiv:2609.37169v1 Announce Type: cross Abstract: Mid-training equips pretrained large language models with specialized and reasoning capabilities, but the returns of this stage are bounded since add...
arXiv AI
1d ago
Alignment via Training Against Probes Without Losing Monitorability
The paper proposes probe-guided fine-tuning, a method that uses probes detecting undesired properties in model activations as a direct training signal. Experiments show that continuously updated probes reduce harmfulness and improve honesty while preserving utility, outperforming DPO and inference-time steering in safety-utility trade-offs and robustness to jailbreak and abliteration attacks. Importantly, the concepts remain linearly encoded after fine-tuning, maintaining monitorability.
arXiv Machine Learning
Aug 6
Looking in the Mirror: Introspecting Side-Effect Misalignments Induced by Fine-Tuning
arXiv:2608. 04347v1 Announce Type: new Abstract: Fine-tuning enables a source model to acquire desired capabilities and behaviors in a target domain while retaining much of its general-purpose competence.
arXiv AI
Sep 10
Generating Pretraining Tokens from Organic Data for Data-Bound Scaling
The paper introduces SynPro, a synthetic data generation framework that augments limited organic text for large language model pretraining by applying rephrasing and reformatting operations. SynPro’s generators are optimized with reinforcement learning rewards for quality, faithfulness, and data influence, and are updated continuously as training plateaus. Experiments on 400M, 1.1B, and 2B models show that SynPro can unlock 3.4–5.2× the effective tokens of standard repetition, even outperforming a non‑data‑bound oracle at larger scales.
Hugging Face Trending Papers
Aug 5
Looking in the Mirror: Introspecting Side-Effect Misalignments Induced by Fine-Tuning
Fine-tuning enables a source model to acquire desired capabilities and behaviors in a target domain while retaining much of its general-purpose competence. However, this adaptation process can also degrade alignment properties that were present in the source model.