Rohan Gupta

Rohan Gupta, Iván Arcuschin, Thomas Kwa et al.

Mechanistic interpretability methods aim to identify the algorithm a neural network implements, but it is difficult to validate such methods when the true algorithm is unknown. This work presents InterpBench, a collection of semi-synthetic yet realistic transformers with known circuits for evaluating these techniques. We train simple neural networks using a stricter version of Interchange Intervention Training (IIT) which we call Strict IIT (SIIT). Like the original, SIIT trains neural networks by aligning their internal computation with a desired high-level causal model, but it also prevents non-circuit nodes from affecting the model's output. We evaluate SIIT on sparse transformers produced by the Tracr tool and find that SIIT models maintain Tracr's original circuit while being more realistic. SIIT can also train transformers with larger circuits, like Indirect Object Identification (IOI). Finally, we use our benchmark to evaluate existing circuit discovery techniques.

Rohan Gupta, Trevor Asbery, Zain Merchant et al.

Coordinating heterogeneous robot fleets to achieve multiple goals is challenging in multi-robot systems. We introduce an open-source and extensible framework for centralized multi-robot task planning and scheduling that leverages LLMs to enable fleets of heterogeneous robots to accomplish multiple tasks. RobotFleet provides abstractions for planning, scheduling, and execution across robots deployed as containerized services to simplify fleet scaling and management. The framework maintains a shared declarative world state and two-way communication for task execution and replanning. By modularizing each layer of the autonomy stack and using LLMs for open-world reasoning, RobotFleet lowers the barrier to building scalable multi-robot systems. The code can be found here: https://github.com/therohangupta/robot-fleet.

Sidhantha Poddar, Rohan Gupta

Script identification and text recognition are some of the major domains in the application of Artificial Intelligence. In this era of digitalization, the use of digital note-taking has become a common practice. Still, conventional methods of using pen and paper is a prominent way of writing. This leads to the classification of scripts based on the method they are obtained. A survey on the current methodologies and state-of-art methods used for processing and identification would prove beneficial for researchers. The aim of this article is to discuss the advancement in the techniques for script pre-processing and text recognition. In India there are twelve prominent Indic scripts, unlike the English language, these scripts have layers of characteristics. Complex characteristics such as similarity in text shape make them difficult to recognize and analyze, thus this requires advance preprocessing methods for their accurate recognition. A sincere attempt is made in this survey to provide a comparison between all algorithms. We hope that this survey would provide insight to a researcher working not only on Indic scripts but also other languages.

Aaron Mueller, Atticus Geiger, Sarah Wiegreffe et al. · stanford

How can we know whether new mechanistic interpretability methods achieve real improvements? In pursuit of lasting evaluation standards, we propose MIB, a Mechanistic Interpretability Benchmark, with two tracks spanning four tasks and five models. MIB favors methods that precisely and concisely recover relevant causal pathways or causal variables in neural language models. The circuit localization track compares methods that locate the model components - and connections between them - most important for performing a task (e.g., attribution patching or information flow routes). The causal variable localization track compares methods that featurize a hidden vector, e.g., sparse autoencoders (SAEs) or distributed alignment search (DAS), and align those features to a task-relevant causal variable. Using MIB, we find that attribution and mask optimization methods perform best on circuit localization. For causal variable localization, we find that the supervised DAS method performs best, while SAE features are not better than neurons, i.e., non-featurized hidden vectors. These findings illustrate that MIB enables meaningful comparisons, and increases our confidence that there has been real progress in the field.

Abrar Anwar, Rohan Gupta, Zain Merchant et al.

Evaluating learned robot control policies to determine their physical task-level capabilities costs experimenter time and effort. The growing number of policies and tasks exacerbates this issue. It is impractical to test every policy on every task multiple times; each trial requires a manual environment reset, and each task change involves re-arranging objects or even changing robots. Naively selecting a random subset of tasks and policies to evaluate is a high-cost solution with unreliable, incomplete results. In this work, we formulate robot evaluation as an active testing problem. We propose to model the distribution of robot performance across all tasks and policies as we sequentially execute experiments. Tasks often share similarities that can reveal potential relationships in policy behavior, and we show that natural language is a useful prior in modeling these relationships between tasks. We then leverage this formulation to reduce the experimenter effort by using a cost-aware expected information gain heuristic to efficiently select informative trials. Our framework accommodates both continuous and discrete performance outcomes. We conduct experiments on existing evaluation data from real robots and simulations. By prioritizing informative trials, our framework reduces the cost of calculating evaluation metrics for robot policies across many tasks.

Rohan Gupta, Erik Jenner

Latent-space monitors aim to detect undesirable behaviours in Large Language Models by leveraging their internal representations rather than relying solely on black-box outputs. These methods have shown promise in identifying behaviours such as deception and unsafe completions. However, these monitors may themselves become training signals, for example, by using problematic samples found in deployment to retrain models. This raises an important question: can models learn to evade such monitors? To evaluate this capability, we introduce RL-Obfuscation, in which LLMs are finetuned via reinforcement learning to evade latent-space monitors while maintaining their blackbox behaviour. We apply RL-Obfuscation to Language Models ranging from 7B to 14B parameters and evaluate their Evasion Success Rate against a suite of monitors. We find that token-level monitors are highly vulnerable to this attack while more holistic monitors, such as max-pooling or attention-based probes, remain robust. Moreover, for these vulnerable monitors, models trained to evade a single static monitor can generalise to evade other unseen monitors. We also find that the models can be trained to conditionally bypass latent-space monitors on only certain inputs. Finally, we study how the models bypass these monitors and find that the model can learn to repurpose tokens to have different internal representations.

Ankur Samanta, Rohan Gupta, Aditi Misra et al.

Molecular property prediction uses molecular structure to infer chemical properties. Chemically interpretable representations that capture meaningful intramolecular interactions enhance the usability and effectiveness of these predictions. However, existing methods often rely on atom-based or rule-based fragment tokenization, which can be chemically suboptimal and lack scalability. We introduce FragmentNet, a graph-to-sequence foundation model with an adaptive, learned tokenizer that decomposes molecular graphs into chemically valid fragments while preserving structural connectivity. FragmentNet integrates VQVAE-GCN for hierarchical fragment embeddings, spatial positional encodings for graph serialization, global molecular descriptors, and a transformer. Pre-trained with Masked Fragment Modeling and fine-tuned on MoleculeNet tasks, FragmentNet outperforms models with similarly scaled architectures and datasets while rivaling larger state-of-the-art models requiring significantly more resources. This novel framework enables adaptive decomposition, serialization, and reconstruction of molecular graphs, facilitating fragment-based editing and visualization of property trends in learned embeddings - a powerful tool for molecular design and optimization.

Abrar Anwar, Rohan Gupta, Jesse Thomason

Robot evaluations in language-guided, real world settings are time-consuming and often sample only a small space of potential instructions across complex scenes. In this work, we introduce contrast sets for robotics as an approach to make small, but specific, perturbations to otherwise independent, identically distributed (i.i.d.) test instances. We investigate the relationship between experimenter effort to carry out an evaluation and the resulting estimated test performance as well as the insights that can be drawn from performance on perturbed instances. We use the relative performance change of different contrast set perturbations to characterize policies at reduced experimenter effort in both a simulated manipulation task and a physical robot vision-and-language navigation task. We encourage the use of contrast set evaluations as a more informative alternative to small scale, i.i.d. demonstrations on physical robots, and as a scalable alternative to industry-scale real world evaluations.

Jay Mundra, Rohan Gupta, Sagnik Mukherjee

This paper describes our contribution to the WASSA 2021 shared task on Empathy Prediction and Emotion Classification. The broad goal of this task was to model an empathy score, a distress score and the overall level of emotion of an essay written in response to a newspaper article associated with harm to someone. We have used the ELECTRA model abundantly and also advanced deep learning approaches like multi-task learning. Additionally, we also leveraged standard machine learning techniques like ensembling. Our system achieves a Pearson Correlation Coefficient of 0.533 on sub-task I and a macro F1 score of 0.5528 on sub-task II. We ranked 1st in Emotion Classification sub-task and 3rd in Empathy Prediction sub-task

Rohan Gupta, Jay Mundra, Deepak Mahajan et al.

In this work, we present our approach for solving the SemEval 2021 Task 2: Multilingual and Cross-lingual Word-in-Context Disambiguation (MCL-WiC). The task is a sentence pair classification problem where the goal is to detect whether a given word common to both the sentences evokes the same meaning. We submit systems for both the settings - Multilingual (the pair's sentences belong to the same language) and Cross-Lingual (the pair's sentences belong to different languages). The training data is provided only in English. Consequently, we employ cross-lingual transfer techniques. Our approach employs fine-tuning pre-trained transformer-based language models, like ELECTRA and ALBERT, for the English task and XLM-R for all other tasks. To improve these systems' performance, we propose adding a signal to the word to be disambiguated and augmenting our data by sentence pair reversal. We further augment the dataset provided to us with WiC, XL-WiC and SemCor 3.0. Using ensembles, we achieve strong performance in the Multilingual task, placing first in the EN-EN and FR-FR sub-tasks. For the Cross-Lingual setting, we employed translate-test methods and a zero-shot method, using our multilingual models, with the latter performing slightly better.