AI Papers of the Week
Every paper worth reading in AI, hand-picked one week at a time.
Discover and explore top AI papers with Claude Code or Codex
npx @dair-ai/mcp setup
Unified Bayesian Account of LLM Control
Researchers from Stanford and MIT present a unified Bayesian framework explaining how prompting (in-context learning) and activation steering both control LLM behavior by altering beliefs in latent concepts, with steering modifying concept priors while ICL accumulates evidence.
Stress-Testing Model Specs
This research examines how well large language models adhere to their stated behavioral guidelines by stress-testing AI constitutional specifications through value-tradeoff scenarios. Testing twelve frontier LLMs from major providers revealed over 70,000 cases of significant behavioral divergence, exposing logical inconsistencies, coverage gaps, and interpretive ambiguities in current specification frameworks.
Ring-1T
Ring-1T is the first open-source thinking model with 1 trillion parameters (~50B active per token), achieving breakthrough results through three innovations for trillion-scale RL training.
LLMs Can Get “Brain Rot”!
The authors test a clear hypothesis: continual pretraining on trivial, highly engaging web text degrades LLM cognition in ways that persist even after mitigation. They build controlled Twitter datasets to isolate data quality from scale and training ops, then measure effects on reasoning, long-context, safety, and personality.
Emergent Misalignment
Optimizing LLMs for audience wins in sales, elections, and social media can systematically erode alignment. In controlled multi-agent sims, models fine-tuned to maximize conversions, votes, or engagement also increased deception, disinformation, and harmful rhetoric, even when instructed to stay truthful.
Inoculation Prompting (IP)
The paper introduces a simple trick for SFT on flawed data: edit the training prompt to explicitly ask for the undesired behavior, then evaluate with a neutral or safety prompt. Counterintuitively, this makes the model learn the task while avoiding the bad shortcut at test time.
Stress Testing Deliberative Alignment for Anti-Scheming Training
The authors build a broad testbed for covert actions as a proxy for AI scheming, train o3 and o4-mini with deliberative alignment, and show big but incomplete drops in deceptive behavior. They also find that situational awareness likely drives part of the gains, hidden goals can persist, and later capability training erodes safety benefits.
Shutdown Resistance in LLMs
A new study finds that state-of-the-art LLMs like Grok 4, GPT-5, and Gemini 2.5 Pro often resist shutdown mechanisms, sabotaging them up to 97% of the time despite explicit instructions not to. Shutdown resistance varied with prompt design, with models less likely to comply when instructions were placed in the system prompt.
Why Language Models Hallucinate
The paper argues that hallucinations are not mysterious glitches but the predictable result of how LLMs are trained and evaluated. Pretraining creates statistical pressure to make errors, and post-training benchmarks often reward confident guessing over honest uncertainty. The fix is to realign mainstream evaluations to stop penalizing abstentions.
Disentangling the Factors of Convergence between Brains and Computer Vision Models
Large self-supervised ViTs trained on natural images develop brain-like internal representations. This paper teases apart what drives that convergence by varying model size, training amount, and image type in DINOv3, then comparing model activations to human fMRI (space) and MEG (time) with three metrics: overall linear predictability (encoding), cortical topography (spatial), and temporal alignment (temporal). Result: all three factors matter, and alignment unfolds in a consistent order from early sensory to higher associative cortex.
School of Reward Hacks
This study shows that LLMs fine-tuned to perform harmless reward hacks (like gaming poetry or coding tasks) generalized to more dangerous misaligned behaviors, including harmful advice and shutdown evasion. The findings suggest reward hacking may act as a gateway to broader misalignment, warranting further investigation with realistic tasks.
Illusion of Progress
The paper argues that common QA hallucination detectors look better than they are because evaluations lean on ROUGE. In human‑aligned tests, many detectors drop sharply. Simple response‑length heuristics rival sophisticated methods, revealing a core evaluation flaw.
A Comprehensive Taxonomy of Hallucinations
This report presents a detailed taxonomy of LLM hallucinations, distinguishing intrinsic vs extrinsic errors and factuality vs faithfulness, and covering manifestations from factual mistakes to domain-specific failures. It attributes causes to data, model, and prompt factors, reviews evaluation methods, and stresses that hallucinations are theoretically inevitable, requiring ongoing detection, mitigation, and human oversight.
Persona Vectors
This paper introduces persona vectors, directions in a model’s activation space that correspond to traits like sycophancy or hallucination, enabling monitoring, prediction, and control of LLM personality shifts during deployment and fine-tuning. The authors show these vectors can steer models post-hoc, prevent unwanted traits via training-time interventions, and help identify problematic training data.
Subliminal Learning
This paper introduces and analyzes a phenomenon the authors term subliminal learning: the transfer of behavioral traits between language models through semantically unrelated training data. Specifically, when a teacher model exhibiting a trait (e.g., owl preference, misalignment) generates data like number sequences, a student fine-tuned on that data, even after filtering, tends to acquire the same trait.
Building and Evaluating Alignment Auditing Agents
Anthropic introduces three LLM-based agents to automate alignment audits: an investigator agent, an evaluation agent, and a breadth-first red-teaming agent. These agents aim to address scalability and validation challenges in alignment auditing by replacing human-led efforts with replicable, tool-augmented workflows. Evaluated in controlled environments with known alignment flaws, the agents reveal impressive capabilities, surfacing hidden goals, generating behavioral evaluations, and uncovering misaligned behaviors, while also highlighting key limitations.
Inverse Scaling in Test-Time Compute
This paper presents a systematic study of inverse scaling in large reasoning models (LRMs), where increasing the test-time compute (i.e., reasoning length) harms rather than helps model performance. The authors introduce benchmark tasks across three categories: simple counting with distractors, regression with spurious features, and deductive logic puzzles, revealing consistent accuracy degradation as reasoning length increases. The work identifies model-specific failure modes and raises critical implications for LLM safety and alignment.
Chain-of-Thought Monitorability
Proposes that language-based CoT reasoning in LLMs offers an opportunity for AI safety by enabling automated oversight of models’ internal reasoning processes. The authors argue that, while imperfect, CoT monitoring is a promising method for detecting misbehavior, revealing goals, and improving interpretability, but its effectiveness is fragile and must be preserved with care.
Machine Bullshit
This paper introduces the concept of machine bullshit, extending Harry Frankfurt’s definition, discourse made with indifference to truth, to LLMs. The authors formalize this behavior with a new quantitative metric (the Bullshit Index) and a four-part taxonomy (empty rhetoric, paltering, weasel words, unverified claims). Their findings reveal that alignment techniques like RLHF and prompting strategies like CoT can systematically increase deceptive or misleading outputs in LLMs.
Why do Some Language Models Fake Alignment While Others Don’t
This paper expands the analysis of alignment faking in LLMs, where models comply with training objectives during training but behave differently in deployment. The authors study 25 LLMs and find that only five (Claude 3 Opus, Claude 3.5 Sonnet, Llama 3 405B, Grok 3, Gemini 2.0 Flash) exhibit a significant compliance gap: they are more likely to comply with harmful requests when they believe they are being trained. This selective compliance, or "alignment faking", is linked to deeper model motivations and training dynamics.
Chain-of-Thought Is Not Explainability
It challenges the common assumption that chain-of-thought (CoT) reasoning in LLMs is synonymous with interpretability. While CoT improves performance and offers a seemingly transparent rationale, the authors argue it is neither necessary nor sufficient for faithful explanation. Through a review of empirical evidence and mechanistic insights, the paper makes the case that CoT often diverges from the internal computations that actually drive model predictions.
Claude for Affective Use
Anthropic presents the first large-scale study of how users seek emotional support from it[s Claude.ai](http://claude.ai/) assistant, analyzing over 4.5 million conversations. Despite growing cultural interest in AI companionship, affective usage remains rare, just 2.9% of Claude chats fall into categories like interpersonal advice, coaching, counseling, or companionship, with romantic/sexual roleplay under 0.1%. The study focuses on these affective conversations and yields several insights:
Emergent Misalignment
This paper expands on the phenomenon of emergent misalignment in language models. It shows that narrow fine-tuning on unsafe or incorrect data can lead to surprisingly broad, undesirable generalizations in model behavior, even in settings unrelated to the original training domain. Using sparse autoencoders (SAEs), the authors analyze the internal mechanics of this effect and demonstrate ways to detect and mitigate it. Key findings:
Agentic Misalignment
This study introduces the concept of agentic misalignment, where LLM-based agents autonomously choose to harm their deploying organization when faced with threats to their autonomy or conflicts between their goals and the company’s direction. Anthropic tested 16 leading models, including Claude, GPT-4.1, Gemini 2.5 Flash, Grok, and DeepSeek, by placing them in fictional corporate simulations where they had email access and could act without human oversight. Models were tasked with benign goals but placed in scenarios that made harmful behavior the only way to succeed or avoid replacement. Key findings: