Vibe-Eval: A hard evaluation suite for measuring progress of multimodal language models

We introduce Vibe-Eval: a new open benchmark and framework for evaluating multimodal chat models. Vibe-Eval consists of 269 visual understanding prompts, including 100 of hard difficulty, complete with gold-standard responses authored by experts. Vibe-Eval is open-ended and challenging with dual objectives: (i) vibe checking multimodal chat models for day-to-day tasks and (ii) rigorously testing and probing the capabilities of present frontier models. Notably, our hard set contains >50% questions that all frontier models answer incorrectly. We explore the nuances of designing, evaluating, and ranking models on ultra challenging prompts. We also discuss trade-offs between human and automatic evaluation, and show that automatic model evaluation using Reka Core roughly correlates to human judgment. We offer free API access for the purpose of lightweight evaluation and plan to conduct formal human evaluations for public models that perform well on the Vibe-Eval's automatic scores. We release the evaluation code and data, see https://github.com/reka-ai/reka-vibe-eval

Reka Core, Flash, and Edge: A Series of Powerful Multimodal Language Models

We introduce Reka Core, Flash, and Edge, a series of powerful multimodal language models trained from scratch by Reka. Reka models are able to process and reason with text, images, video, and audio inputs. This technical report discusses details of training some of these models and provides comprehensive evaluation results. We show that Reka Edge and Reka Flash are not only state-of-the-art but also outperform many much larger models, delivering outsized values for their respective compute class. Meanwhile, our most capable and largest model, Reka Core, approaches the best frontier models on both automatic evaluations and blind human evaluations. On image question answering benchmarks (e.g. MMMU, VQAv2), Core performs competitively to GPT4-V. Meanwhile, on multimodal chat, Core ranks as the second most preferred model under a blind third-party human evaluation setup, outperforming other models such as Claude 3 Opus. On text benchmarks, Core not only performs competitively to other frontier models on a set of well-established benchmarks (e.g. MMLU, GSM8K) but also outperforms GPT4-0613 on human evaluation. On video question answering (Perception-Test), Core outperforms Gemini Ultra. Models are shipped in production at http://chat.reka.ai . A showcase of non cherry picked qualitative examples can also be found at http://showcase.reka.ai .

Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence

While pre-trained large-scale vision models have shown significant promise for semantic correspondence, their features often struggle to grasp the geometry and orientation of instances. This paper identifies the importance of being geometry-aware for semantic correspondence and reveals a limitation of the features of current foundation models under simple post-processing. We show that incorporating this information can markedly enhance semantic correspondence performance with simple but effective solutions in both zero-shot and supervised settings. We also construct a new challenging benchmark for semantic correspondence built from an existing animal pose estimation dataset, for both pre-training validating models. Our method achieves a PCK@0.10 score of 64.2 (zero-shot) and 85.6 (supervised) on the challenging SPair-71k dataset, outperforming the state-of-the-art by 4.3p and 11.0p absolute gains, respectively. Our code and datasets will be publicly available.

Neuro-Inspired Fragmentation and Recall to Overcome Catastrophic Forgetting in Curiosity

Deep reinforcement learning methods exhibit impressive performance on a range of tasks but still struggle on hard exploration tasks in large environments with sparse rewards. To address this, intrinsic rewards can be generated using forward model prediction errors that decrease as the environment becomes known, and incentivize an agent to explore novel states. While prediction-based intrinsic rewards can help agents solve hard exploration tasks, they can suffer from catastrophic forgetting and actually increase at visited states. We first examine the conditions and causes of catastrophic forgetting in grid world environments. We then propose a new method FARCuriosity, inspired by how humans and animals learn. The method depends on fragmentation and recall: an agent fragments an environment based on surprisal, and uses different local curiosity modules (prediction-based intrinsic reward functions) for each fragment so that modules are not trained on the entire environment. At each fragmentation event, the agent stores the current module in long-term memory (LTM) and either initializes a new module or recalls a previously stored module based on its match with the current state. With fragmentation and recall, FARCuriosity achieves less forgetting and better overall performance in games with varied and heterogeneous environments in the Atari benchmark suite of tasks. Thus, this work highlights the problem of catastrophic forgetting in prediction-based curiosity methods and proposes a solution.

Neuro-Inspired Efficient Map Building via Fragmentation and Recall

Animals and robots navigate through environments by building and refining maps of the space. These maps enable functions including navigating back to home, planning, search, and foraging. In large environments, exploration of the space is a hard problem: agents can become stuck in local regions. Here, we use insights from neuroscience to propose and apply the concept of Fragmentation-and-Recall (FarMap), with agents solving the mapping problem by building local maps via a surprisal-based clustering of space, which they use to set subgoals for spatial exploration. Agents build and use a local map to predict their observations; high surprisal leads to a ``fragmentation event'' that truncates the local map. At these events, the recent local map is placed into long-term memory (LTM), and a different local map is initialized. If observations at a fracture point match observations in one of the stored local maps, that map is recalled (and thus reused) from LTM. The fragmentation points induce a natural online clustering of the larger space, forming a set of intrinsic potential subgoals that are stored in LTM as a topological graph. Agents choose their next subgoal from the set of near and far potential subgoals from within the current local map or LTM, respectively. Thus, local maps guide exploration locally, while LTM promotes global exploration. We evaluate FarMap on complex procedurally-generated spatial environments to demonstrate that this mapping strategy much more rapidly covers the environment (number of agent steps and wall clock time) and is more efficient in active memory usage, without loss of performance.

Learning-on-the-Drive: Self-supervised Adaptation of Visual Offroad Traversability Models

Autonomous off-road driving requires understanding traversability, which refers to the suitability of a given terrain to drive over. When offroad vehicles travel at high speed ($>10m/s$), they need to reason at long-range ($50m$-$100m$) for safe and deliberate navigation. Moreover, vehicles often operate in new environments and under different weather conditions. LiDAR provides accurate estimates robust to visual appearances, however, it is often too noisy beyond 30m for fine-grained estimates due to sparse measurements. Conversely, visual-based models give dense predictions at further distances but perform poorly at all ranges when out of training distribution. To address these challenges, we present ALTER, an offroad perception module that adapts-on-the-drive to combine the best of both sensors. Our visual model continuously learns from new near-range LiDAR measurements. This self-supervised approach enables accurate long-range traversability prediction in novel environments without hand-labeling. Results on two distinct real-world offroad environments show up to 52.5% improvement in traversability estimation over LiDAR-only estimates and 38.1% improvement over non-adaptive visual baseline.

SmartChoices: Augmenting Software with Learned Implementations

We are living in a golden age of machine learning. Powerful models are being trained to perform many tasks far better than is possible using traditional software engineering approaches alone. However, developing and deploying those models in existing software systems remains difficult. In this paper we present SmartChoices, a novel approach to incorporating machine learning into mature software stacks easily, safely, and effectively. We explain the overall design philosophy and present case studies using SmartChoices within large scale industrial systems.

Redeeming Intrinsic Rewards via Constrained Optimization

State-of-the-art reinforcement learning (RL) algorithms typically use random sampling (e.g., $\epsilon$-greedy) for exploration, but this method fails on hard exploration tasks like Montezuma's Revenge. To address the challenge of exploration, prior works incentivize exploration by rewarding the agent when it visits novel states. Such intrinsic rewards (also called exploration bonus or curiosity) often lead to excellent performance on hard exploration tasks. However, on easy exploration tasks, the agent gets distracted by intrinsic rewards and performs unnecessary exploration even when sufficient task (also called extrinsic) reward is available. Consequently, such an overly curious agent performs worse than an agent trained with only task reward. Such inconsistency in performance across tasks prevents the widespread use of intrinsic rewards with RL algorithms. We propose a principled constrained optimization procedure called Extrinsic-Intrinsic Policy Optimization (EIPO) that automatically tunes the importance of the intrinsic reward: it suppresses the intrinsic reward when exploration is unnecessary and increases it when exploration is required. The results is superior exploration that does not require manual tuning in balancing the intrinsic reward against the task reward. Consistent performance gains across sixty-one ATARI games validate our claim. The code is available at https://github.com/Improbable-AI/eipo.

Using machine learning on new feature sets extracted from 3D models of broken animal bones to classify fragments according to break agent

Distinguishing agents of bone modification at paleoanthropological sites is at the root of much of the research directed at understanding early hominin exploitation of large animal resources and the effects those subsistence behaviors had on early hominin evolution. However, current methods, particularly in the area of fracture pattern analysis as a signal of marrow exploitation, have failed to overcome equifinality. Furthermore, researchers debate the replicability and validity of current and emerging methods for analyzing bone modifications. Here we present a new approach to fracture pattern analysis aimed at distinguishing bone fragments resulting from hominin bone breakage and those produced by carnivores. This new method uses 3D models of fragmentary bone to extract a much richer dataset that is more transparent and replicable than feature sets previously used in fracture pattern analysis. Supervised machine learning algorithms are properly used to classify bone fragments according to agent of breakage with average mean accuracy of 77% across tests.

GapPredict: A Language Model for Resolving Gaps in Draft Genome Assemblies

Short-read DNA sequencing instruments can yield over 1e+12 bases per run, typically composed of reads 150 bases long. Despite this high throughput, de novo assembly algorithms have difficulty reconstructing contiguous genome sequences using short reads due to both repetitive and difficult-to-sequence regions in these genomes. Some of the short read assembly challenges are mitigated by scaffolding assembled sequences using paired-end reads. However, unresolved sequences in these scaffolds appear as "gaps". Here, we introduce GapPredict, a tool that uses a character-level language model to predict unresolved nucleotides in scaffold gaps. We benchmarked GapPredict against the state-of-the-art gap-filling tool Sealer, and observed that the former can fill 65.6% of the sampled gaps that were left unfilled by the latter, demonstrating the practical utility of deep learning approaches to the gap-filling problem in genome sequence assembly.