We introduce KiloGram, a resource for studying abstract visual reasoning in humans and machines. Drawing on the history of tangram puzzles as stimuli in cognitive science, we build a richly annotated dataset that, with >1k distinct stimuli, is orders of magnitude larger and more diverse than prior resources. It is both visually and linguistically richer, moving beyond whole shape descriptions to include segmentation maps and part labels. We use this resource to evaluate the abstract visual reasoning capacities of recent multi-modal models. We observe that pre-trained weights demonstrate limited abstract reasoning, which dramatically improves with fine-tuning. We also observe that explicitly describing parts aids abstract reasoning for both humans and models, especially when jointly encoding the linguistic and visual inputs. KiloGram is available at https://lil.nlp.cornell.edu/kilogram .

While the capabilities of autonomous systems have been steadily improving in recent years, these systems still struggle to rapidly explore previously unknown environments without the aid of GPS-assisted navigation. The DARPA Subterranean (SubT) Challenge aimed to fast track the development of autonomous exploration systems by evaluating their performance in real-world underground search-and-rescue scenarios. Subterranean environments present a plethora of challenges for robotic systems, such as limited communications, complex topology, visually-degraded sensing, and harsh terrain. The presented solution enables long-term autonomy with minimal human supervision by combining a powerful and independent single-agent autonomy stack, with higher level mission management operating over a flexible mesh network. The autonomy suite deployed on quadruped and wheeled robots was fully independent, freeing the human supervision to loosely supervise the mission and make high-impact strategic decisions. We also discuss lessons learned from fielding our system at the SubT Final Event, relating to vehicle versatility, system adaptability, and re-configurable communications.

Effective management of public shared spaces such as car parking space, is one challenging transformational aspect for many cities, especially in the developing World. By leveraging sensing technologies, cloud computing, and Artificial Intelligence, Cities are increasingly being managed smartly. Smart Cities not only bring convenience to City dwellers, but also improve their quality of life as advocated for by United Nations in the 2030 Sustainable Development Goal on Sustainable Cities and Communities. Through integration of Internet of Things and Cloud Computing, this paper presents a successful proof-of-concept implementation of a framework for managing public car parking spaces. Reservation of parking slots is done through a cloud-hosted application, while access to and out of the parking slot is enabled through Radio Frequency Identification (RFID) technology which in real-time, accordingly triggers update of the parking slot availability in the cloud-hosted database. This framework could bring considerable convenience to City dwellers since motorists only have to drive to a parking space when sure of a vacant parking slot, an important stride towards realization of sustainable smart cities and communities.

Random graph models with community structure have been studied extensively in the literature. For both the problems of detecting and recovering community structure, an interesting landscape of statistical and computational phase transitions has emerged. A natural unanswered question is: might it be possible to infer properties of the community structure (for instance, the number and sizes of communities) even in situations where actually finding those communities is believed to be computationally hard? We show the answer is no. In particular, we consider certain hypothesis testing problems between models with different community structures, and we show (in the low-degree polynomial framework) that testing between two options is as hard as finding the communities. In addition, our methods give the first computational lower bounds for testing between two different `planted' distributions, whereas previous results have considered testing between a planted distribution and an i.i.d. `null' distribution.

Despite recent success in large language model (LLM) reasoning, LLMs still struggle with hierarchical multi-step reasoning like generating complex programs. In these cases, humans often start with a high-level algorithmic design and implement each part gradually. We introduce Parsel, a framework enabling automatic implementation and validation of complex algorithms with code LLMs, based on hierarchical function descriptions in natural language. Parsel can be used across domains requiring hierarchical reasoning, e.g. code synthesis, theorem proving, and robotic planning. We demonstrate Parsel's capabilities by using it to generate complex programs that cannot currently be automatically implemented from one description and backtranslating Python programs in the APPS dataset. Beyond modeling capabilities, Parsel allows problem-solving with high-level algorithmic designs, benefiting both students and professional programmers.

Current image generation models struggle to reliably produce well-formed visual text. In this paper, we investigate a key contributing factor: popular text-to-image models lack character-level input features, making it much harder to predict a word's visual makeup as a series of glyphs. To quantify the extent of this effect, we conduct a series of controlled experiments comparing character-aware vs. character-blind text encoders. In the text-only domain, we find that character-aware models provide large gains on a novel spelling task (WikiSpell). Transferring these learnings onto the visual domain, we train a suite of image generation models, and show that character-aware variants outperform their character-blind counterparts across a range of novel text rendering tasks (our DrawText benchmark). Our models set a much higher state-of-the-art on visual spelling, with 30+ point accuracy gains over competitors on rare words, despite training on far fewer examples.

Large "instruction-tuned" language models (finetuned to respond to instructions) have demonstrated a remarkable ability to generalize zero-shot to new tasks. Nevertheless, they depend heavily on human-written instruction data that is limited in quantity, diversity, and creativity, therefore hindering the generality of the tuned model. We introduce Self-Instruct, a framework for improving the instruction-following capabilities of pretrained language models by bootstrapping off its own generations. Our pipeline generates instruction, input, and output samples from a language model, then prunes them before using them to finetune the original model. Applying our method to vanilla GPT3, we demonstrate a 33% absolute improvement over the original model on Super-NaturalInstructions, on par with the performance of InstructGPT_001, which is trained with private user data and human annotations. For further evaluation, we curate a set of expert-written instructions for novel tasks, and show through human evaluation that tuning GPT3 with Self-Instruct outperforms using existing public instruction datasets by a large margin, leaving only a 5% absolute gap behind InstructGPT_001. Self-Instruct provides an almost annotation-free method for aligning pre-trained language models with instructions, and we release our large synthetic dataset to facilitate future studies on instruction tuning.

Transformers have been essential to pretraining success in NLP. Other architectures have been used, but require attention layers to match benchmark accuracy. This work explores pretraining without attention. We test recently developed routing layers based on state-space models (SSM) and model architectures based on multiplicative gating. Used together these modeling choices have a large impact on pretraining accuracy. Empirically the proposed Bidirectional Gated SSM (BiGS) replicates BERT pretraining results without attention and can be extended to long-form pretraining of 4096 tokens without approximation.

Language models have recently achieved strong performance across a wide range of NLP benchmarks. However, unlike benchmarks, real world tasks are often poorly specified, and agents must deduce the user's intended behavior from a combination of context, instructions, and examples. We investigate how both humans and models behave in the face of such task ambiguity by proposing AmbiBench, a new benchmark of six ambiguously-specified classification tasks. We evaluate humans and models on AmbiBench by seeing how well they identify the intended task using 1) instructions with varying degrees of ambiguity, and 2) different numbers of labeled examples. We find that the combination of model scaling (to 175B parameters) and training with human feedback data enables models to approach or exceed the accuracy of human participants across tasks, but that either one alone is not sufficient. In addition, we show how to dramatically improve the accuracy of language models trained without large-scale human feedback training by finetuning on a small number of ambiguous in-context examples, providing a promising direction for teaching models to generalize well in the face of ambiguity.

We introduce INSTRUCTOR, a new method for computing text embeddings given task instructions: every text input is embedded together with instructions explaining the use case (e.g., task and domain descriptions). Unlike encoders from prior work that are more specialized, INSTRUCTOR is a single embedder that can generate text embeddings tailored to different downstream tasks and domains, without any further training. We first annotate instructions for 330 diverse tasks and train INSTRUCTOR on this multitask mixture with a contrastive loss. We evaluate INSTRUCTOR on 70 embedding evaluation tasks (66 of which are unseen during training), ranging from classification and information retrieval to semantic textual similarity and text generation evaluation. INSTRUCTOR, while having an order of magnitude fewer parameters than the previous best model, achieves state-of-the-art performance, with an average improvement of 3.4% compared to the previous best results on the 70 diverse datasets. Our analysis suggests that INSTRUCTOR is robust to changes in instructions, and that instruction finetuning mitigates the challenge of training a single model on diverse datasets.