REVIEW 12 cited by
Language Models as Agent Models
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
Language Models as Agent Models
read the original abstract
Language models (LMs) are trained on collections of documents, written by individual human agents to achieve specific goals in an outside world. During training, LMs have access only to text of these documents, with no direct evidence of the internal states of the agents that produced them -- a fact often used to argue that LMs are incapable of modeling goal-directed aspects of human language production and comprehension. Can LMs trained on text learn anything at all about the relationship between language and use? I argue that LMs are models of intentional communication in a specific, narrow sense. When performing next word prediction given a textual context, an LM can infer and represent properties of an agent likely to have produced that context. These representations can in turn influence subsequent LM generation in the same way that agents' communicative intentions influence their language. I survey findings from the recent literature showing that -- even in today's non-robust and error-prone models -- LMs infer and use representations of fine-grained communicative intentions and more abstract beliefs and goals. Despite the limited nature of their training data, they can thus serve as building blocks for systems that communicate and act intentionally.
Forward citations
Cited by 12 Pith papers
-
Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolution
Promptbreeder evolves both task prompts and the mutation prompts that improve them using LLMs, outperforming Chain-of-Thought and Plan-and-Solve on arithmetic and commonsense reasoning benchmarks.
-
Emergent Misalignment Recruits a Pre-existing Persona Subspace
Fine-tuning on narrow bad data recruits a low-rank persona subspace already present in a frozen instruction-tuned model; holding that subspace out of activations prevents broad misalignment, and injecting it into the ...
-
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.
-
On the Limits of Prompt-Conditioned Language Models as General-Purpose Learners
Prompt-conditioned LLMs face irreducible error floors from language's limited information capacity and alignment constraints, proven via PAC-Bayes bounds on bilevel cheap-talk games for certain task families.
-
Procedural Knowledge Is Not Low-Rank: Why LoRA Fails to Internalize Multi-Step Procedures
On multi-step procedural tasks, LoRA fine-tuning underperforms full fine-tuning at every rank tested because procedural knowledge requires high-rank weight updates.
-
Probing Persona-Dependent Preferences in Language Models
Linear probes on residual-stream activations extract a preference vector that tracks and steers pairwise task choices across personas in Gemma-3-27B and Qwen-3.5-122B, including anti-correlated evil personas.
-
Probing Persona-Dependent Preferences in Language Models
Linear probes on residual-stream activations identify a shared preference vector in LLMs that tracks choices across prompts and causally steers decisions even for anti-correlated personas.
-
Steering grids for sparse-autoencoder features: when a top-context label names an activation regime rather than a causal axis
Pairwise matrices for SAEs demonstrate that single-feature inspection mislabels causal axes, with joint suppression and matched-geometry controls revealing distinct output regimes not captured by single-feature or ran...
-
Steering grids for sparse-autoencoder features: when a top-context label names an activation regime rather than a causal axis
A steering-grid protocol that varies coefficient and joint feature set shows that standard single-feature SAE labels are often incomplete, naming an activation regime rather than a causal axis.
-
Compiling Agentic Workflows into LLM Weights: Near-Frontier Quality at Two Orders of Magnitude Less Cost
Compiling agentic workflows into LLM weights creates subterranean agents with near-frontier quality at two orders of magnitude less cost, validated empirically on travel booking, Zoom support, and insurance claims tasks.
-
Framing Effects in Independent-Agent Large Language Models: A Cross-Family Behavioral Analysis
Prompt framing significantly shifts LLM choices toward risk-averse options in a threshold voting task even when the prompts are logically equivalent.
-
Psychological Imagination Networks Show Cross-Population Centrality and Clustering Alignment in Humans That Large Language Models Fail to Replicate
Human vividness-rating networks are correlated across populations and cluster by questionnaire context, whereas LLM-derived networks are mostly degenerate single-clusters, showing a human-LLM divergence in imagined-sc...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.