Mapping Language Models to Grounded Conceptual Spaces

By Roma Patel, Ellie Pavlick. 2022

208 citations recorded by the source corpus.

Subfields: Large Language Models, Cognitive Science & Psychology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Large Language Models

In Cognitive Science & Psychology

Citations in the corpus

Emergent Abilities of Large Language Models (2022, neutral): “Figure 2F shows the task of grounded conceptual mappings, where language models must learn to map a conceptual domain, such as a cardinal direction, represented in a textual grid world (Patel & Pavlick, 2022).”
Neural Theory-of-Mind? On the Limits of Social Intelligence in Large LMs (2022, background): “, 2021), which can be tied to real-world grounding with a small amount of additional supervised training (Patel and Pavlick, 2022).”
Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks (2023, background): “, color, size) often plausibly mirror those of the grounded world (Abdou et al., 2021; Patel and Pavlick, 2022; Mollo and Millière, 2023).”
The alignment problem from a deep learning perspective (2022, background): “Large neural networks can learn robust concepts, including concepts corresponding to high-level environmental outcomes [Patel and Pavlick, 2022, Jaderberg et al., 2019, Meng et al., 2022].”
Mechanistic Interpretability for AI Safety - A Review (2024, background): “Additionally, as mechanistic interpretability focuses on causal understanding, novel methods encompass variants of activation patching for uncovering causal mechanisms and causal scrubbing for hypothesis evaluation.”
Can Language Models Solve Graph Problems in Natural Language? (2023, background): “Builda-Graph prompting encourages LLMs to map the textual descriptions of graphs and structures to grounded conceptual spaces [Patel and Pavlick, 2022] before tackling the specific problem through a one-sentence instruction, while Algorithmic prompting instructs LLMs to revisit the algorithmic steps”
Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task (2022, background): “Recent evidence suggests language models can develop internal representations for very simple concepts, such as color, direction Abdou et al. (2021); Patel & Pavlick (2022), or tracking boolean states during synthetic tasks (Li et al., 2021) (see Related Work (section 6) for more detail).”
Language Models Represent Space and Time (2023, background): “Abdou et al. (2021) and Patel & Pavlick (2021) show LLMs have representations that reflect perceptual and conceptual structure in color and spatial domains.”
Grounding Large Language Models in Interactive Environments with Online Reinforcement Learning (2023, methodology): “We reuse PPO’s hyperparameters from Ramamurthy et al. [2022] and did not perform any further tuning (see Table 7).”
The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasets (2023, background): “Substantial previous work has studied whether LLMs encode world models in their representations (Li et al., 2023a; 2021; Abdou et al., 2021; Patel & Pavlick, 2022).”
ProgPrompt: Generating Situated Robot Task Plans using Large Language Models (2022, background): “Prompting LLMs to generate text useful for robot task planning is a nascent topic [32], [33], [34], [2], [4],”