Agent Evals: Finally, With The Map

Key Takeaways

hi this is Ari heluk from root signals uh presenting agent evaluations finally with the map so the agent evaluation is rather Art and Science and ultimately it is nonetheless required to actually ensure that your agents do what you expect them to do especially when you're launching them in production so without further Ado let's just uh Dive Right There into the map and start to get an understanding of what are all the things involved if you want to evaluate all the aspects of your agents uh so the agent evaluation can be actually neatly divided to the semantic and behavioral parts of what the agents do the semantic part is all about how do the representations uh of the reality that the agent has actually relate to the reality whereas the behavioral parts are all about how do the actions and the tools that the agent is using actually contributing to the achievement of the agent's goals in its environment and ultimately what kind of effects it will have on its environment as well the semantic part is uh further divided into what could be call the single step or single turn uh items which are like coherence consistency Etc so various sematic virtues and what can be called the multi- turn aspect of the same so this Rel to chatter reasoning whereas in the behavioral part you also have this distinction whether you're looking at the task progression and planning or then individual selection and usage of tools so let's uh dive into each of these and paying also attention to the fact that truthfulness is ultimately achiev by grounding uh the representations in your data often with track whereas uh the goal achievement is what actually uh is achieved by grounding to the tools that the agent has available and the this is sort of a symmetry that is not accidental because in a sense there's similarity and analogy between uh representations and behaviors because representing the world is a kind of activity so you can see that uh that the representations are in a sense a special case of tools special case of uh uh of behaviors and uh let's then dive into these individual parts of the map so first looking at the semantic quality for the single turn case so here we have these Universal virtues there's a long list of these that you can look at uh and these are actually non agenic in a sense and we are only covering them here because of completeness so that you can sort of understand these agent Parts through the contrast so these are things like uh uh is the uh the reply that the agent is giving to the user is this consistent is the is the content actually safe and so on and so on and then there's the inter interesting and important part of whether whatever the agent is actually saying if it's uh aligning uh with the values of uh of the organizations or people uh who are the stakeholders and also whether it aderes to the policies of the same rag or what could be more generally called yet attention management uh is then something that you need to measure through specific evaluators such as those looking at whether the retreat context uh was correct whether all of it was comprehensively recalled uh and so on and so on and ultimately relating the answers uh to the external reality through what is often called the Fai fulness which is then separate from uh from the both the answer question relevance and also from the notion of factfulness in general which relates to the reality Beyond just the reference data that the rag is using so now moving forward uh we should also pay attention to the fact that the rag evaluations actually come uh in many forms and they they also have certain symmetries uh that are not that difficult to understand when you look at them uh through uh a map like this so they are all just essentially looking at different relationships between the parts of the rag pipeline but this is not the the main topic of the presentation so you can look look up more information through the uh background material links that people provide so then moving to the multi-turn case the multi-turn in the semantic quality sense means essentially chat conversation histories how do these develop what sort of things we want to look for in this consistency Etc adherence and sticking to the topics when that is actually necessary sometimes we want to allow changing topics if the for example a chatbot users user wants to change the topic or sometimes we don't but we need to be able to uh be aware of this then super important uh and almost groundbreaking progress has been achieved by uh by uh investments in reasoning so you can also uh actually evaluate reasoning traces the reasoning uh Chain of Thought if if allowed to use that term here and that's sort of another way of looking at the kind of sequential or M well multi- turn and sequential kinds of activities that the agent is just doing in the course of its reasoning and representations of the world before we even taking any kind of actions and now we move to the side so here we have uh things to evaluate such as where the agent is actually following instructions is it extracting the the tool characteristics correctly is it selecting the right is the right tool being selected is the output quality of the tools correct are the error situations handled correctly are the structures interation tool format are they uh correct so these are all all the things that relate to uh to the behaviors even before you have a chain of uh behaviors and when you move to this chain of behaviors or multistep or uh multi-turn uh cases so then you actually start to look at like are the actions that the ACT agent is taking converging towards actually achieving its goal and uh is the plan that the agent might have actually consistent and is it actually uh high quality in whatever sense you want to measure that uh so then both of these are actually grounded uh on external reality so uh like mentioned the representations are grounded on truthfulness and uh the activities and behaviors are ultimate grounded by goal achievement and utility uh and this is uh what sort of uh is the ultimate uh metric to everything else and know uh that the agent is trying to do and all these other things along the way are sort of just proxy metrics so to speak uh moving to other practical considerations we can only scratch the surface here so I'm just sort of listing not to leave this out but they don't needly fall into this general map that I presented but should also be taken into account on the background so most important of which is the cost and latency optimization so generally you want to the agent to progress towards its goal as quickly as possible and as cheaply as possible so cost and latency optimization the number of steps optimization and then moving to the what you ofer are going to need which is the tracing and debugging so being able to actually see where the agent go wrong error management here refers to to dealing with the with the errors of the tool usage so not the actual like semantic errors that that the agent is doing in the course of its inference and thinking process uh very important distinction to keep in mind is the offline versus online testing so what are the things that uh that you can actually uh try out and evaluate during development and what are the things that you can do and must do or should do uh during the actual online activities of the agent and these are actually going to become like a two very distinct Dimensions that could be also included on the map but uh but that would make the map map more complicated so that's why I'm just mentioning them here now separately and then we have various special cases that I don't have time to cover some of these uh are uh more refined and more advanced than others and and some of these things are sort of uh more researchy and there's a lot of papers on each of these topics and uh uh depending on what you're doing tool specific metrics might actually be useful to add to the mix even uh uh and they might even be rather simple to implement because the tools are often something as easy as straightforward as like API calls and uh and such things that can actually be just measured separately uh using more traditional software testing methodologies uh one important part to mention here is that a lot of these measurements are going to to be implemented with llm as charge techniques and now the caveat uh behind this is that the quite often when people start implementing these evaluation methodologies they are looking at this kind of a single tier approach single tier here means that you're focusing on optimizing this operative element flow meaning that you're optimizing your agent which makes sense but if you only are concerned of uh with getting some scores on what the agent is doing then you're sort of forgetting that what about all the cost latencies uncertainty related to the charge itself the charge that is on the background so you should be actually uh taking into account as early as possible that you are going to need to also optimize the charge itself and this is what the what we could call Double tier so you need to optimize both the operative LM flow that powers your agent and then this chargement flow that actually Powers your evalu ations and uh this is a rather complex situation in general and we are calling this eval Ops because uh this seems like a like a separate kind of thing that involves uh evaluations that themselves are so complicated uh so expensive and slow that they sort of earn their own uh category of of activities and this is uh something that uh that we have written about and uh and you can also find more about this on the source materials uh the general general tast of it is that that eval Ops is a kind of a special case of llm Ops and and actually operates on different kinds of entities that that the llm Ops uh uh in general and uh and then it also requires different ways of thinking different kind of software implementations and also essentially a different kind of resourcing to to get those evaluations right so thank you very much this was only a brief Glimpse on the on the general landscape I hope the map is helpful to you and uh please take a look at the source materials which will give you more depth to each of these uh these topics and uh happy to discuss any of these things and let me uh know if I'm forgetting something crucial and uh this uh presentation will of course be obsolete by the time it it goes out so uh probably when when you're watching this there's already some major developments have happen but that's uh that's how it is on this General Life of uh of uh AI Engineers uh so let's go out there and make our agents measurable controllable and let's make sure they are actually doing our biding and not not rebelling against our uh ultimate intentions thank you very much